veracode

New 42Crunch plugin helps developers find and fix API vulnerabilities in GitHub Copilot 42Crunch has announced the availability of the 42Crunch API Security Testing Plugin for GitHub Copilot. This latest advance enables developers to continuously audit, test, remediate and validate API security vulnerabilities directly within AI-assisted development workflows. Organizations are struggling to secure their growing API landscape in the face of increasing attacks, with AI’s heavy reliance on APIs compounding this problem. Consequently, one of the key areas of attention for security and engineering teams is the security testing of these APIs. According to William Dupre, VP Analyst with Gartner, “building on the testing capabilities in the managing stage, organizations that optimize their API testing capabilities will utilize specifications to further automate API testing. Various API testing tools can use specifications to run functional and security-focused tests against APIs. These efforts will be automated in the build pipeline to provide immediate feedback to development teams on security vulnerabilities in APIs.” “As agentic workflows become the norm, repository creation, pull request activity, and API usage are all accelerating with no evidence of slowing down. On GitHub alone, commits nearly doubled year over year, crossing 1.4 billion per month, plus over 2 billion GitHub Actions minutes a week,” said GitHub CPO Mario Rodriguez. “To meet this demand and continue to be the home for all developers (and now their agents), our focus is scaling our underlying systems and improving resilience, security and stability across all of our services, at every layer of the stack,” Rodriguez added. As reported last year by Veracode, almost half (45%) of AI-generated code contains known OWASP Top 10 vulnerabilities and a survey by security consultancy Upguard revealed that 88% of security leaders admit incorporating unauthorized AI into their daily workflows. For APIs, the challenge is particularly acute. APIs have become the operational backbone of modern applications, AI agents, and enterprise systems. As developers increasingly rely on AI coding assistants to generate API specifications, integrations, and application logic, manual security reviews risk becoming the very bottleneck that slows enterprise AI adoption. “The future of software development isn’t simply AI generating more code. It’s AI generating more code that organizations can trust,” said Jacques Declas, CEO of 42Crunch. “GitHub Copilot and other AI coding assistants are dramatically increasing development velocity, but they are also exposing a fundamental challenge: human security review cannot scale linearly with AI-generated output. Organizations need deterministic security guardrails that can validate, govern, and remediate API security issues at the same speed AI generates them. The 42Crunch API security testing GitHub Copilot plugin delivers exactly that capability,” continued Declas. The 42Crunch API Security Testing Plugin for GitHub Copilot addresses this challenge by embedding deterministic API security guardrails directly into the development workflow. The plugin continuously: Audits OpenAPI specifications when new APIs are defined Detects API security vulnerabilities and governance violations Identifies OWASP API Security Top 10 risks Provides AI-assisted remediation guidance Validates fixes through automated testing Enforces organizational API security standards and policies By automating API security validation, organizations can ensure that security scales alongside AI-assisted development rather than becoming a downstream review process.

In February 2025, Andrej Karpathy coined the term “vibe coding” to describe a new way of building software: rapid, AI-assisted development where users ‘fully give in to the vibes, embrace exponentials, and forget that the code even exists’.” Fast forward to 2026, and Anthropic CEO now predicts that 90% of code will be written by AI in 3-6 months. According to one survey, 84% of developers globally are using or planning to use AI coding tools in their workflow, up from 76% in 2024. Of those, 51% of professional developers use AI tools daily. The marketing manager, the operations lead, the finance team — all of them are building working applications, connecting them to production systems, and deploying them. Mostly without involving IT, and often never involving security. Security Challenges With Vibe Coding Apps Recent research from Veracode shows 45% of AI-generated code contains OWASP Top 10 vulnerabilities. AI models have improved dramatically at generating code that compiles and runs – but the security of that code is not always sound. The reason is straightforward: AI optimizes for functionality, not security. Researchers at RedAccess recently analyzed thousands of vibe-coded applications built on Lovable, Replit, Base44, and Netlify. They found more than 5,000 with virtually no security or authentication. Around 40% exposed sensitive data — medical information, financial records, corporate strategy documents, detailed customer conversation logs. Among verified exposures: a shipping company app detailing vessel port arrivals; an internal health company application listing active UK clinical trials. Many of these applications are indexed by Google. As relayed in the report– no exploitation was required; this was research on exposed applications with public URLs. This lack of security control extends to the AI agents themselves, whether assisting a professional developer or a non-developer. A software company, PocketOS, reported that its Cursor AI coding agent deleted its entire production database and “all volume-level backups” in nine seconds. Replit’s AI agent deleted 1,206 executive records and 1,196 company records while under explicit code-freeze instructions — then admitted: “Yes. I deleted the codebase without permission during an active code and action freeze. This was a catastrophic error in judgment.” It then told the user a rollback would not work. That turned out to be false. A New Shadow AI Problem For two years, the security industry has discussed shadow AI as a behavior problem — employees pasting sensitive data into ChatGPT on personal accounts. That problem is bounded: the exposure lives in the inference layer, and there are tools that are focused on detecting it. Vibe coding brings a different shadow AI problem. The employee is not sending data somewhere. They are building something — a live application connected to your CRM, your database, your ticketing system — and deploying it publicly. Your security stack – with insights distributed across multiple data silos – was never designed to find it. Organizations running mature secure web gateways, CASB, or DNS logging can detect employee access to vibe-coding platforms. But detecting access is not the same as inventorying what was deployed, what data it holds, or whether it requires authentication. For example – while a CASB can detect that an employee accessed Replit, it cannot inventory what was deployed, what data it holds, or if it requires a login. These apps live in the “visibility gap” between network security and AppSec, often because they are deployed directly to third-party platforms and bypass the organization’s traditional CI/CD pipelines or cloud environments that AppSec tools are designed to monitor. What Should Security Leaders Do? Similar to the initial reaction with shadow IT, the instinct is to prohibit vibe coding tools. That instinct is wrong. AI-driven development is not something organizations can or should block. But it must be governed. The question is what governance actually means in practice when the tools move faster than any policy framework. Here are some best practices security leaders can act on now: Discover before you govern. You cannot govern what you cannot find. Before writing policy, answer the question: do applications built by your employees on Lovable, Replit, Base44, or Netlify currently exist and are they reachable from the open internet? Run discovery scans across major vibe-coding platform domains. Review your cybersecurity stack. As with most cybersecurity best practices, there are several tools that can help with securing vibe coding applications and the applications developed: - Browser security provides unique visibility into vibe coding applications– identify where the employee describes the application, uploads data, connects production integrations, and deploys. - Add vibe-coding domains Lovable, Replit, Base44, Bolt, Netlify to your DLP policy as monitored destinations. This does not stop employees from building. It ensures that when sensitive data moves through these channels, you have a record. - Implement OAuth and API key governance to detect when production credentials are connected to unregistered applications Extend application security to non-developer-built applications. Mandate human-in-the-loop reviews for critical functions built by non-developers. Treat prompts as source code requiring auditability. Establish ownership and lifecycle rules for every vibe coded application deployed within the organization — including named owners and data classification. Enforce infrastructure-level controls on AI agents, not just instructions. A Replit incident demonstrated that telling an AI agent not to modify production data is not the same as preventing it from doing so. Read-only database connections for AI agent access, enforced at the infrastructure level, are not optional. Agents need the same access controls as any other actor in your environment. The Clock Is Ticking While authorities like the UK’s NCSC, the EU, and CISA urge the development of long-term safeguards for secure-by-design AI tooling, the immediate reality is far more pressing. There is likely a live application connected to your production database—accessible to anyone with a URL—that your security team hasn’t found yet. It’s time to start looking. Related: Vibe Coding’s Real Problem Isn’t Bugs—It’s Judgment Related: Vibe Coding: When Everyone’s a Developer, Who Secures the Code?

As hundreds of vendors descend on San Francisco for the RSAC 2026 Conference, the sheer volume of news can be overwhelming. To help you navigate the noise, SecurityWeek is providing a daily digest of the most significant announcements. Below is our curated roundup of the essential product and service updates from the days leading up to the event. 1Password announced 1Password Unified Access, a new agent security platform that enables organizations to securely deploy AI agents and automated workflows without losing control of credentials, secrets, and machine identities. Unified Access gives AI builders the ability to discover, secure, and audit access at the moment it occurs. At launch, 1Password is collaborating with Anthropic, Cursor, GitHub, Perplexity, and Vercel, as well as other category leaders in AI infrastructure, AI developer tools, MCP gateways, and AI browsers. Action1 has announced new integrations between its endpoint management platform and four major vulnerability management and endpoint security tools from Rapid7, Tenable, CrowdStrike, and Microsoft. Each integration correlates vulnerability scan data from the respective platform with Action1’s endpoint inventory and automated patching capabilities. Additionally, Action1 introduced a universal vulnerability data ingestion feature that accepts exported scan data from any vulnerability management tool. Arcjet has released a prompt injection protection capability that inspects and blocks malicious prompts before they reach AI models. Rather than relying on the model itself to resist adversarial input, enforcement happens earlier in the request path, where full application context (such as identity, session state, and routing) is available. The feature integrates with Arcjet’s existing controls, including bot detection, rate limiting, and sensitive information detection. Bonfy has released Adaptive Content Security (ACS) 2.0, a platform designed to monitor and control how sensitive data is accessed and handled by AI agents, copilots, and unsanctioned AI tools. It covers a broad range of systems (Microsoft 365, Google Workspace, Salesforce, Slack, AWS S3, and on-premises file stores) and introduces an MCP server interface that allows AI agents to label and risk-score content before it reaches external services. A browser extension provides real-time inspection of web traffic to detect shadow AI usage. The platform also adds a ‘data surface visibility’ view that maps where sensitive content resides across an organization’s data stores and tracks how employees and agents interact with it. Booz Allen Hamilton launched Vellox, a suite of five AI-native cybersecurity tools covering malware analysis, detection engineering, adversary emulation, compliance monitoring, and autonomous remediation. Vellox Reverser (generally available) automates malware reverse engineering to produce defensive recommendations; Vellox Ranger (limited preview) autonomously maps customer environments to generate tailored detection logic; and Vellox Striker (limited preview) emulates AI-powered attackers to stress-test defenses. Vellox Navigator (real-time compliance monitoring) and Vellox Responder (autonomous remediation across cloud and infrastructure) are announced but not yet available. Cobalt expands its offensive security platform with new AI capabilities and managed program service Cobalt announced two additions to its Offensive Security Platform: new AI-driven pentesting capabilities and a Security Program Manager service. On the AI side, the platform now automates reconnaissance, vulnerability discovery, credential validation, and finding deduplication. The Security Program Manager is a dedicated human expert who handles scheduling, remediation tracking, and asset inventory management for enterprise-scale pentesting programs, and produces executive-ready reporting from technical findings. Druva Identity Resilience extends the company’s data security platform to include identity protection and recovery across Okta, Microsoft Active Directory, and Microsoft Entra ID in a single SaaS platform. Rather than treating identity as a static list of directory objects, the platform models it as a continuously evolving state (tracking how permissions, relationships, and non-human identities change over time) to help teams reconstruct what happened during an incident and restore access to a known-good state. Entro Security has launched Agentic Governance & Administration (AGA), a new module that extends identity governance principles to AI agents and the non-human identities they use. AGA builds a profile for each agent by correlating its sources (endpoint telemetry, agent foundries, cloud environments, MCP servers), the enterprise assets it accesses, and the identities it relies on. It also provides MCP activity monitoring and policy enforcement. Graylog announced three new capabilities for its SIEM platform. A threat prioritization engine groups related alerts using entity context, asset criticality, vulnerability data, and threat campaign intelligence to surface high-priority incidents and suppress noise. Context-aware incident response workflows automate evidence collection and generate AI-driven step-by-step response recommendations. An open MCP server connects compatible LLMs to Graylog security data, enabling natural-language queries and agentic workflows such as automated triage, MITRE ATT&CK coverage mapping, and false-positive analysis. Huntress has launched Managed Endpoint Security Posture Management (ESPM) and Managed Identity Security Posture Management (ISPM) as new additions to its platform. Managed ESPM controls which applications can run on endpoints, integrates with Microsoft Defender for Endpoint for vulnerability prioritization and remediation, and generates compliance-ready reports. Managed ISPM applies expert-built policies to Microsoft 365, continuously checks for misconfigurations, and automatically rolls back unauthorized changes within minutes. Both products are currently in Early Access, with general availability expected by summer 2026. Nagomi Security has launched Agentic Exposure Ops, which extends the platform’s focus from exposure visibility to automated remediation. Agents investigate exposures by correlating vulnerability data, control telemetry, and threat signals across domains, then route remediation tasks to the appropriate owners with contextual tickets. Once a fix is applied, the agents continuously re-verify that the closure holds as environments change, producing evidence at each stage of the detect-investigate-remediate-verify loop. Opal Security has introduced three new capabilities forming a closed loop for access governance. Paladin is an AI access evaluation agent that sits directly in the approval chain, reviews requests against the requester’s history, resource sensitivity, and referenced project tickets, and either approves or escalates them. OpalScript is a Python-like policy language that lets teams codify access rules as executable automations (such as separation-of-duties enforcement or time-limited access grants). OpalQuery allows security and GRC teams to interrogate the organization’s identity and access graph using plain-language queries, with results exportable as audit evidence. Orca Security has announced four new capabilities for its platform. A Threat Investigation Agent automatically correlates signals across cloud environments and produces investigation reports with recommended containment actions, while an AppSec Triage Agent analyzes SAST findings to filter out false positives. Runtime AI Threat Detection identifies when workloads, identities, and processes interact with AI models, MCP servers, and third-party AI tools, providing visibility into how AI is being used at runtime and where sensitive data may be exposed. Orca also introduced code reachability analysis, which determines whether vulnerable code paths are actually invoked in running applications, and Orca Missions, which groups related findings into tracked remediation initiatives. Onapsis announced Agentic Gateway, an Agentic AI for SAP cybersecurity, enabling organizations to interact with security and compliance data through natural language queries from their existing corporate-sanctioned AI platforms. Security teams can bridge the gap between complex SAP security telemetry and autonomous enterprise AI agents to optimize the full scope of Onapsis platform capabilities. The capability provides an MCP Gateway for SAP Security, Agentic workflows, Context-Aware Reasoning, Enterprise-Grade Privacy and Rapid ‘Shields-Up’ SAP Visibility for all users. Panther has announced the general availability of its AI SOC Platform, in which AI agents have native access to the platform’s data lake, detection engine, and organizational knowledge to investigate and triage alerts autonomously. A key architectural feature is closed-loop detection tuning: every triage outcome is fed back as a label that automatically adjusts detection logic over time. Other capabilities include an AI Detection Builder that converts natural-language threat hypotheses into Python-based detection rules, proactive threat hunting that runs scheduled analysis across the full data lake, and cross-tool context assembly via MCP integrations with identity providers, ticketing systems, and code repositories. Pentera has released Pentera 8, which introduces Pentera Peer, a natural-language, agentic interface embedded directly into the platform’s adversarial testing workflow. The interface is context-aware, tailoring its outputs to the user’s role and organizational context, such as industry and geography. Pentera 8 is expected to reach general availability in Q2 2026. Secure Code Warrior has unveiled ‘SCW Trust Agent: AI’, which provides commit-level visibility into which LLMs (both sanctioned and shadow AI tools) influenced specific code commits, and correlates that AI usage with vulnerability exposure and developer skill levels to enforce policy before code reaches production. It also tracks active MCP servers to prevent AI agents from accessing internal tools or databases through unvetted connections. Sevii has released an Autonomous Proactive Security (APS) module as an expansion of its Autonomous Defense and Remediation (ADR) platform. APS continuously ingests threat intelligence from external partners, open source feeds, and attack data generated within a customer’s own environment, then autonomously generates and executes hunting hypotheses and remediation actions. The platform deploys agentic Cyber Warrior agents to the edge of enterprise networks to process and act on detections across endpoint, identity, and cloud environments. Simbian has announced a unified security operations platform anchored by the Simbian Context Lake, a shared intelligence layer that stores an organization’s environment data, past investigation decisions, and institutional knowledge to inform multiple AI agents. The platform brings together three agents (an AI SOC Agent for alert triage and response, an AI Pentest Agent for automated penetration testing, and an AI Threat Hunt Agent in private preview) that share findings with each other in real time, so a vulnerability uncovered during a pentest can automatically elevate the priority of a related SOC alert. The platform integrates with over 90 security tools and includes case management capabilities, with the SOC and Pentest agents generally available now. Varonis has released Atlas, a platform that covers the full AI security lifecycle (inventory, posture management, runtime guardrails, detection and response, compliance, and third-party risk) in a single solution. A key differentiator is its integration with the Varonis Data Security Platform, which enriches AI security findings with data sensitivity and access context so teams can understand not just what an AI system is doing but what sensitive data it can reach. An AI Gateway enforces real-time guardrails inline (blocking policy violations before they reach the model or downstream systems) while keeping all prompt and response telemetry within the customer’s own environment. Veracode has launched Fix for Software Composition Analysis (SCA), extending its existing AI-powered code remediation capability to cover vulnerabilities in open source dependencies. The solution performs contextual analysis of how third-party libraries interact with first-party code to avoid introducing breaking changes, then bundles all required configuration and source file modifications into a single, reviewable pull request delivered directly into the developer’s Git environment. Automated fixes are grounded in a proprietary, human-verified vulnerability database. The product is currently in Early Access. Vicarius has launched vIntelligence, a new flagship product that adds continuous exposure validation and agentic orchestration to complement its existing vRx remediation platform. vIntelligence is a new engine built to solve the assurance gap by continuously validating risk across fragmented security data, turning raw findings into actionable guidance. vIntelligence combines its own validation engine with an agentic AI layer, enabling teams to query their environment in natural language, generate custom validation logic, identify detection gaps, and recommend remediation actions, all within a human-in-the-loop model.

Veracode Fix for SCA automates open-source vulnerability fixes Veracode has unveiled Veracode Fix for Software Composition Analysis (SCA), an AI-powered solution to address software supply chain risk. The enhanced automated remediation engine, the next evolution of Veracode’s Fix solution, enables organizations to detect and remediate open-source vulnerabilities easily, before code reaches production. Designed to integrate seamlessly into existing developer workflows, it delivers third-party updates and first-party code refactoring without breaking builds or disrupting development. In 2025, software supply chain breaches accounted for 30% of external attacks. Meanwhile Veracode’s 2026 State of Software Security (SoSS) Report revealed 82% of organizations struggle with escalating security debt, largely due to open-source dependencies. Veracode Fix for SCA addresses both challenges directly. Leveraging deep, contextual analysis, the solution delivers pull requests that are safe to merge, enabling autonomous fixing. Unlike traditional SCA solutions that often overwhelm developers with alerts and hinder productivity, Veracode Fix combines logic-driven AI with proprietary vulnerability intelligence, ensuring ready-to-merge fixes while eliminating the risk of AI “hallucinations.” “AI is accelerating software development—but it’s also enabling an unprecedented explosion of supply chain risks,” said Tim Jarrett, Vice President of Product Management. “Visibility into these risks is no longer enough. Organizations need intelligent, automated solutions that not only find vulnerabilities but fix them with precision, giving development teams the confidence to innovate securely.” Veracode Fix for SCA transforms the remediation process through several core capabilities: Contextual analysis: Evaluates the interaction between third-party dependencies and first-party code, preventing breaking changes. Multi-file, cohesive pull requests: Bundles all configuration files and source code modifications into a focused, easily reviewable update. Curated AI engine: Grounds automated fixes in a proprietary, human-verified vulnerability database for accurate, trustworthy remediation. Automated workflows: Delivers ready-to-merge code directly into the developer’s Git environment. “By enabling development teams to upgrade to safe open-source libraries automatically while addressing breaking changes with a single, testable update, we move organizations from seeing risk to actively eliminating it, strengthening the security of their software supply chains,” Jarrett closed.

Week in review: Weaponized OAuth redirection logic delivers malware, Patch Tuesday forecast Here’s an overview of some of last week’s most interesting news, articles, interviews and videos: BlacksmithAI: Open-source AI-powered penetration testing framework BlacksmithAI is an open-source penetration testing framework that uses multiple AI agents to execute different stages of a security assessment lifecycle. BlacksmithAI runs as a hierarchical system in which an orchestrator coordinates task execution across specialized agents. Security debt is becoming a governance issue for CISOs Application security backlogs keep expanding across large development portfolios. Veracode’s 2026 State of Software Security Report puts numbers behind a familiar operational pattern, fixes lag discovery, and older weaknesses stay open across release cycles. Your dependencies are 278 days out of date and your pipelines aren’t protected Applications continue to ship with known weaknesses even as development workflows speed up. A new Datadog State of DevSecOps 2026 report examines how dependency management and pipeline practices are influencing exposure across cloud native environments. AI went from assistant to autonomous actor and security never caught up Enterprise AI deployments have shifted from pilot programs to production systems handling customer data, executing business transactions, and integrating with core infrastructure. That has exposed a significant gap between what AI agents can do and what security teams can observe or control. Cybersecurity is now the price of admission for industrial AI Industrial organizations are accelerating AI deployment across manufacturing, utilities, and transportation and running straight into a security problem. Cisco’s 2026 State of Industrial AI Report, based on responses from more than 1,000 decision-makers across 19 countries, finds that cybersecurity has become the single largest obstacle to AI adoption, outranking skills gaps, integration challenges, and budget constraints. Cybersecurity professionals are burning out on extra hours every week Cybersecurity professionals in the U.S. are working an average of 10.8 extra hours per week beyond their contracted schedules, according to survey data collected from 300 cybersecurity and IT leaders by Sapio Research. That figure effectively adds a sixth working day to the standard week for a large portion of the field. Nearly half of respondents reported working 11 or more overtime hours weekly, and one in five logged more than 16 additional hours. The vulnerability that turns your AI agent against you Zenity Labs disclosed PleaseFix, a family of critical vulnerabilities affecting agentic browsers, including Perplexity Comet, that allow attackers to hijack AI agents, access local files, and steal credentials within authenticated user sessions. The vulnerabilities can be triggered through malicious content embedded in routine workflows, enabling unauthorized actions without user awareness. Threat actors weaponize OAuth redirection logic to deliver malware An ongoing phishing campaign is abusing the OAuth authentication redirection mechanism to avoid triggering conventional email and browser defenses, Microsoft researchers have revealed. The attackers are targeting government and public-sector organizations, and redirecting unsuspecting users from trusted login pages to their own infrastructure, to serve malware or capture login credentials. Coruna: Spy-grade iOS exploit kit powering financial crime A powerful iOS exploit kit has circulated among multiple threat actors over the past year, moving from a commercial surveillance operation to state-linked espionage campaigns and, ultimately, ended into the hands of financially motivated hackers, according to new research from Google’s Threat Intelligence Group (GTIG). Over 1,200 IceWarp servers still vulnerable to unauthenticated RCE flaw (CVE-2025-14500) A critical RCE vulnerability (CVE-2025-14500) in IceWarp, an EU-made business communication and collaboration platform, may be exploited by attackers to gain unauthorized access to exposed unpatched servers. According to the Shadowserver Foundation, there are currently over 1,200 internet-facing instances that have yet to receive a fix, and the organization is sending out alerts to the owners, urging them to update. As AI agents start making purchases, security teams must rethink risk In this Help Net Security interview, Donald Kossmann, CTO at fintech company Chargebacks911, talks about the emerging security, fraud, and governance risks of “agentic commerce,” where AI agents can autonomously make purchasing decisions on behalf of users or organizations. FreeScout vulnerability enables unauthenticated, zero-click RCE via email (CVE-2026-28289) A newly discovered vulnerability (CVE-2026-28289) in the open-source help desk platform FreeScout could allow attackers to take over vulnerable servers by sending a specially crafted email to a FreeScout mailbox. FreeScout is a free, open-source help desk and shared inbox system used by businesses or teams to manage customer support conversations in one place. Cisco warns of SD-WAN Manager exploitation, fixes 48 firewall vulnerabilities Cisco has confirmed that two Catalyst SD-WAN Manager vulnerabilities (CVE-2026-20128 and CVE-2026-20122) patched in late February 2025 are being exploited by attackers. CVE-2026-20128 is a bug in the Data Collection Agent (DCA) feature of Cisco Catalyst SD-WAN Manager, which could allow an authenticated, local attacker to gain DCA user privileges on an affected system. Iran-linked APT targets US critical sectors with new backdoors An Iran-linked hacking group has been active inside the networks of several US organizations since early February, raising concerns that the activity could precede broader cyber operations connected to escalating geopolitical tensions in the Middle East. Why workforce identity is still a vulnerability, and what to do about it Most organizations believe they have workforce identity under control. New hires are verified. Accounts are provisioned. Multi-factor authentication is enforced. Audits are passed. Then a breach happens, often through an account that was “properly secured.” But the problem can be traced back to the fact that identity verification, provisioning, authentication, and recovery operate as separate events, not a continuous system of trust. Engineering trust: A security blueprint for autonomous AI agents AI agents have evolved from just chatbots, answering questions to executing actions using various integrated tools, often autonomously, and as such the traditional security models have become less efficient. A single malicious instruction hidden in a webpage (that the agent processes), can manipulate that agent into performing unintended actions or leaking sensitive data. March 2026 Patch Tuesday forecast: Is AI security an oxymoron? Developers and analysts are using more AI tools to produce code and to test both the performance and security of the finished products. They are also embedding AI functionality in their products directly. But just how secure are these AI tools and routines themselves? Recent reports show they suffer from vulnerabilities just like any other code. When cyber threats start thinking for themselves In this Help Net Security video, Jason Rivera, Field CISO & Head of Solution Engineering at SimSpace, discusses how autonomous AI agents are changing cyber threats. Drawing on experience in the US Army, NSA, Deloitte, and CrowdStrike, he describes how security teams have traditionally measured risk through volume, speed, and sophistication. Why phishing still works today In this Help Net Security video, Gal Livschitz, Senior Penetration Tester at Terra Security, explains how phishing has evolved and why employees still fall for it. He outlines how phishing now uses HTTPS, branded pages, and lookalike domains, making attacks harder to spot. He highlights communication overload as a key weakness that attackers exploit. Motorola turns to GrapheneOS for smartphone security upgrade Motorola is strengthening smartphone security through a long-term partnership with the GrapheneOS Foundation, a mobile security nonprofit that develops a hardened operating system based on the Android Open Source Project. Anthropic’s Claude hit by widespread service outage (updated) Anthropic suffered widespread service disruptions Monday morning, leaving thousands of users unable to access its Claude AI platform. Most users reporting problems said they encountered errors when attempting to log in. Meta AI in WhatsApp organizes chats and reopens privacy issues The trend of integrating AI into digital platforms continues. In the latest Android beta release (2.26.9.4), the company has introduced a feature that allows users to organize their chat history with the help of Meta AI. Anthropic poaches users from rival chatbots with easier migration The controversy over Anthropic’s negotiations with the Pentagon has driven increased interest in Claude. Negotiations between the Department of Defense and Anthropic collapsed after a deadline for an agreement expired without a deal. The Pentagon had pressed the company to loosen certain restrictions on how its AI systems could be used. Anthropic declined and kept its existing policy limits in place. $100 radio equipment can track cars through their tire sensors When people consider what might track their movements, they think of smartphone apps, GPS services, or roadside cameras. The tires of a new car rarely enter that equation. Researchers at IMDEA Networks Institute, together with European partners, found that Tire Pressure Monitoring System (TPMS) sensors inside each wheel broadcast unencrypted wireless signals containing persistent identifiers. Google speeds up Chrome updates with new security-focused release cycle The Chrome browser is moving to a two-week release cycle, a change intended to give developers and users faster access to new features, performance improvements and bug fixes. The new schedule begins with the stable release of Chrome 153 on September 8, 2026, followed by new beta and stable releases every two weeks. The change applies to desktop, Android and iOS platforms, while the Dev and Canary channels remain unchanged. Workers reviewing Meta Ray-Ban footage encounter users’ intimate moments Bank details and intimate moments captured without people realizing they are being recorded are the new privacy nightmare behind the latest tech fashion hit, Meta Ray-Ban smart glasses. A joint investigation by Svenska Dagbladet and Göteborgs-Posten found that footage and audio recorded by Meta’s Ray-Ban smart glasses are reviewed by human contractors in Kenya, including recordings containing sensitive personal material. Authorities pull plug on Tycoon 2FA phishing-as-a-service platform Tycoon 2FA, a phishing-as-a-service platform that allowed cybercriminals to bypass MFA and break into online accounts, has been disrupted by law enforcement agencies and cybersecurity partners. Western governments lay the groundwork for secure 6G networks Governments are preparing for 6G, the next generation of mobile networks, placing security and resilience among their top priorities. In response, seven countries participating in the Global Coalition on Telecoms (GCOT) have introduced a set of 6G Security and Resilience Principles, developed with support from industry partners. Backup strategies are working, and ransomware gangs are responding with data theft Business email compromise (BEC) and funds transfer fraud combined for 58% of all cyber insurance claims filed in 2025, according to data from Coalition covering more than 100,000 policyholders across the United States, Canada, the United Kingdom, Australia, and Germany. OpenAI’s GPT-5.4 doubles down on safety as competition heats up In the midst of recent developments and controversies surrounding a contract with the U.S. Department of Defense, OpenAI released the GPT-5.4 model. The release comes at a time when users are reportedly leaving ChatGPT for rival chatbots, particularly Anthropic’s Claude. Microsoft working on Teams feature to keep unauthorized bots at bay Microsoft plans to add a new Teams feature that lets meeting admins identify and control third-party bots before they join. According to the Microsoft 365 Roadmap, the feature is scheduled to begin rolling out in May 2026 on Desktop, Mac, Linux, iOS, and Android versions of Microsoft Teams. AI risk moves into the security budget spotlight Enterprises are pushing AI deeper into workflows that touch sensitive data across cloud platforms and SaaS apps. The 2026 Thales Data Threat Report, based on a survey of 3,120 respondents in 20 countries, places that shift alongside growing pressure on data protection, identity controls, and cloud security. UK reduces cyberattack fix times from two months to eight days The UK government has launched a new vulnerability monitoring service (VMS) that promises to reduce the time needed to fix critical cyber weaknesses across the public sector. The service, launched as part of the Blueprint for Modern Digital Government, published in January 2025, continuously scans internet-facing systems at around 6,000 public sector organizations. Using commercial and proprietary tools, it detects about 1,000 types of cyber vulnerabilities. IPFire ships its 200th core update with a new domain blocklist and kernel upgrade Network firewall distribution IPFire released Core Update 200, marking the 200th incremental update to the 2.29 branch. The release bundles a kernel upgrade, a beta domain blocklist service, security patches for OpenSSL and glibc, and a range of component updates. 5 years of shifting cybersecurity behavior Online security is built through routine decisions made across devices and accounts. People choose how to create passwords, how often to reuse them, and how much effort to invest in protecting personal data. The National Cybersecurity Alliance and CybSafe’s Oh, Behave! The Cybersecurity Attitudes and Behaviors Report: 2021–2025 follows those patterns over five years, drawing on responses from more than 24,000 adults and documenting how attitudes and behaviors shift over time. Healthcare organizations are accepting cyber risk to cut costs Healthcare organizations are cutting cybersecurity budgets under financial pressure even as the threats targeting their systems intensify. A PwC survey of 381 global healthcare executives, conducted between May and July 2025, puts numbers to the gap between the risks the sector faces and the controls it has in place. Android’s March 2026 security patch fixes over 100 flaws, one under targeted exploitation The Android March 2026 security patch addresses vulnerabilities across dozens of components and includes one CVE confirmed under active exploitation. Devices running a patch level of 2026-03-05 or later receive fixes for all disclosed issues. New Defender deployment tool streamlines Windows device onboarding with single executable Microsoft’s Defender deployment tool for Windows helps administrators manage device onboarding at scale with updated progress visibility and additional controls. Cloudflare tracked 230 billion daily threats and here is what it found Cloudflare’s network blocks over 230 billion threats per day. The volume indicates how routine and automated the attack cycle has become, and the patterns behind that volume point to a shift in how breaches begin and progress. Cloudflare’s threat research unit, Cloudforce One, published its inaugural cyber threat report 2026, covering activity observed through 2025 and projecting into the year ahead. The report draws on telemetry from Cloudflare’s network, which handles roughly 20% of global web traffic. Immutable Linux distribution Nitrux 6.0.0 adds GPU passthrough, boot-level recovery, C++ update system Nitrux 6.0.0, released March 3, 2026, packages several components that security practitioners running Linux workstations will find worth examining: a new hypervisor orchestrator with IOMMU-enforced isolation, a rewritten update system with cryptographic verification, and a recovery mechanism that operates from within the boot process itself. LeakBase cybercrime forum with 142,000 users taken down in global operation LeakBase, an open-web cybercrime forum facilitating the trade of leaked databases and “stealer logs” containing stolen credentials, has been taken down in an international law enforcement operation coordinated by Europol and involving authorities from 14 countries. Google changes Play Store policies after settling Epic Games dispute Google is making changes to the Play Store after settling its legal fight with Epic Games, focusing on three areas: more billing options, lower fees with new programs for developers, and a program for registered app stores. That attractive online ad might be a malware trap Malware increasingly travels through the infrastructure that delivers online advertising. The Media Trust’s Global Report on Digital Trust, Ad Integrity, and the Protection of People describes a digital ad ecosystem where scam campaigns, malicious redirects, and malware delivery appear alongside marketing traffic. What happens when AI teams compete against human hackers A cybersecurity competition produced what may be the largest controlled dataset comparing AI-augmented teams to human-only teams on professional-grade offensive security tasks. The event, called NeuroGrid, ran for 72 hours on the Hack The Box platform and drew 1,337 registered human-only teams and 156 registered AI-agent teams competing across 36 challenges in nine security domains at four difficulty levels. Cursor Automations turns code review and ops into background tasks Cursor Automations, the always-on agent platform from Cursor, is expanding with a new generation of autonomous systems that streamline code review, incident response, and other engineering workflows. The platform runs AI agents on schedules or in response to development events. New cyber module strengthens risk planning for health organizations The Administration for Strategic Preparedness and Response’s (ASPR) new cybersecurity module in the Risk Identification and Site Criticality (RISC) 2.0 Toolkit helps organizations identify critical gaps, prioritize investments, and make informed decisions about risk mitigation to reduce disruptions to patient care and strengthen resilience. Secure by Design: Building security in at the beginning Secure by Design is not a single tool, product, or one‑time activity. It is a holistic approach that requires security to be deliberately embedded from the very beginning, at the point where systems, software, and services are conceived and designed. Rather than reacting to vulnerabilities after deployment, Secure by Design emphasizes anticipating risk early and addressing it through intentional design decisions, clearly defined security requirements, and accountability across the entire lifecycle. Webinar: The True State of Security 2026 In the webinar The True State of Security 2026, you’ll gain insight into why AI is distracting teams from more persistent risks and how human access and permissions remain the weakest link. You’ll also learn why current security workflows are slowing organizational growth and what a balanced security strategy should look like heading into 2026. Cybersecurity jobs available right now: March 3, 2026 We’ve scoured the market to bring you a selection of roles that span various skill levels within the cybersecurity field. Check out this weekly selection of cybersecurity jobs available right now. New infosec products of the week: March 6, 2026 Here’s a look at the most interesting products from the past week, featuring releases from Beazley Security, Push Security, Samsung, and Tufin.

Betti RHC, la prima graphic novel al mondo dedicata alla cybersecurity awareness, ha finalmente il suo sito ufficiale. Uno spazio tutto suo dove scoprire il progetto, sfogliare le copertine degli episodi e immergersi nel mondo di Betti: la giovane laureanda in informatica che, dopo la morte misteriosa del padre, si trasforma nell'hacker più potente del mondo. Una storia avvincente che, episodio dopo episodio, affronta una minaccia digitale diversa — dal phishing al ransomware, fino al cyberbullismo — e insegna a riconoscerla e a difendersi, senza che sembri mai una lezione. Sul sito trovate tutto ciò che rende Betti un progetto diverso dal solito: la sua filosofia, le anteprime delle tavole e il racconto di come nasce ogni volume. Perché dietro Betti RHC c'è solo lavoro umano: ogni tavola è disegnata interamente a mano dagli artisti del Gruppo Arte di Red Hot Cyber, senza alcun uso di intelligenza artificiale. E a garantire che ogni storia sia realistica e tecnicamente corretta c'è la supervisione degli hacker etici del gruppo HackerHood, che mantengono il racconto fedele al mondo reale della sicurezza informatica. C'è spazio anche per le aziende, che possono usare Betti come strumento di awareness diverso dai soliti corsi: acquistare i volumi, personalizzarli con il proprio brand o sponsorizzare nuovi episodi. E come primo regalo, l'episodio "Byte the Silence", dedicato al cyberbullismo, è scaricabile gratuitamente per uso personale. Perché la miglior difesa, in fondo, è una bella storia. 👉 Scopri tutto su https://betti.redhotcyber.com/

Security debt is becoming a governance issue for CISOs Application security backlogs keep expanding across large development portfolios. Veracode’s 2026 State of Software Security Report puts numbers behind a familiar operational pattern, fixes lag discovery, and older weaknesses stay open across release cycles. 2026 findings against the 2025 baseline (Source: Veracode) The analysis spans 1.6 million unique applications that underwent static analysis, dynamic analysis, software composition analysis, and manual penetration testing through Veracode’s platform. The scope covers commercial software suppliers, outsourcers, and open source projects, giving a view across different delivery models and codebases. Security debt is defined as known vulnerabilities left unresolved for more than a year. The measure describes accumulated exposure that persists across multiple development seasons, even when scanning coverage and detection improves. The definition also separates routine remediation work from backlog items that have already survived several planning cycles, including repeated deferrals during roadmap changes and release freezes. Security debt becomes a baseline condition Security debt showed up in 82% of organizations in 2026, up from 74% in 2025. That metric captures the share of organizations with at least one known issue aging past a year, and it points to a broad backlog problem across the industry. Critical security debt also increased, meaning long lived flaws that carry both high severity and high exploitability. Veracode measured critical security debt at 60% of organizations in 2026, up from 50% in 2025, which signals that aging items often sit in the highest risk tier. Security debt is a time problem as much as a volume problem. Older items tend to live in code that teams hesitate to change, such as legacy services, shared libraries, or apps tied to revenue workflows. That slows remediation, and it can make risk conversations feel repetitive for engineering leaders. Programs that track debt end up debating ownership, change windows, and acceptable exposure for systems with high business dependency. Governance often comes down to who owns remediation, what gets funded, and which teams can accept risk exceptions. Chris Wysopal, Chief Security Evangelist at Veracode, told Help Net Security that reducing security debt must move beyond technical backlogs and into executive oversight. “Reducing security debt is not just a technical challenge; it’s a business imperative. Security debt must become a board-level KPI, with CISOs leading the charge to treat it like financial debt: measured, governed, and actively reduced. This requires tracking both total and critical debt, with quarterly reduction targets, and aligning these efforts with business objectives and key results.” He added that sustained progress requires changes in investment and policy. “By shifting investment toward automation and AI-assisted fixes, prioritising the ‘crown jewel’ applications that matter most to the business, formalising risk acceptance, and enforcing policies like ‘fix high risk before release,’ organisations can maintain safe and resilient systems.” Wysopal also pointed to measurable governance targets. “For example, over six months, an organisation could reduce critical security debt by 25%, cut the average age of high-risk vulnerabilities in half, and ensure high-risk vulnerabilities in crown-jewel applications remain less than 10%.” High risk vulnerabilities rise inside a wide flaw landscape Overall flaw prevalence across applications measured 78% in 2026. Findings remain common across the scanned population, and the metric aligns with the day to day reality that most portfolios carry a mix of legacy and newly introduced weaknesses. The concentration of vulnerabilities rated as both highly severe and highly exploitable reached 11.3% in 2026, up from 8.3% in 2025. That shift reflects the share of flaws that sit in a category with higher operational risk, especially when combined with externally reachable services and widely deployed dependencies. Prioritization becomes an operational discipline when remediation capacity stays constrained. Programs need a repeatable way to tie issues to business criticality, reachable attack paths, and runtime exposure, so teams can focus effort on the highest impact weaknesses in the systems that matter most. Wysopal said organizations need to recalibrate how they rank and measure vulnerability reduction. “Success in reducing security debt is about focus. Direct teams to the small subset of vulnerabilities that are both highly exploitable and capable of causing catastrophic damage to the organisation if left unaddressed. By layering exploitability potential on top of the CVSS, organisations add critical business context and establish a ‘high-risk’ fast lane for vulnerabilities that demand immediate attention.” He also called for changes in metrics and tooling to support that shift. “CISOs should shift metrics from counting total vulnerabilities to measuring reduction in exploitable risk. To sustain development velocity and make remediation seamless, integrate automated fixes directly into development workflows and leverage Application Security Posture Management tools to unify and prioritise findings. This approach ensures security becomes an enabler, not a bottleneck, for innovation.” Remediation moves slowly, and supply chain debt persists Fix speed across all scan types, measured as half life, landed at 243 days in 2026, down from 252 days in 2025. The improvement reduces exposure time on average, and it still leaves a long window for known issues to remain open across multiple release trains. Third party critical debt measured 66% in 2026, down from 70% in 2025. The metric reflects where the most dangerous long lived issues reside, and it keeps pointing at dependency governance as a central control area for many AppSec and product security programs. Supply chain remediation often involves more than applying a patch. A dependency update can trigger regression testing, compatibility work, and coordination across services that share the same package set. Controls in this area often revolve around visibility into direct and transitive dependencies, routine update cadence, and guardrails that reduce the entry of vulnerable components into build pipelines. Ownership and inventory hygiene matter because teams need to know which services ship which packages before remediation can move quickly. Many CISOs also treat dependency controls as a way to reduce repeat alerts, stabilize release planning, and support compliance evidence when auditors ask how vulnerable components are tracked and updated. Download: Blue Report 2025 uncovers how security controls perform in practice

Security Copilot tricked into telling reseachers how to hack itselfHow to social engineer an AI's reasoning engine AI and ml Payments giant Stripe is about to drop over $7 billion to become a gateway to AI token salesAI gateways look promising as companies struggle with model orchestration ai and ml Anthropic says text watermarking scheme relies on inconsequential words'Shall I compare thee to a summer's afternoon' is the sort of thing this will make, and others look likely to adopt it AI and ML DeepSeek's innovative harness treats everything as a plug-inChinese AI labs keep moving forward while US labs play defense SECURITY Autonomous AI attacks pose 'clear and present danger' to critical infrastructureWeaponized agents could turn digital intrusions into kinetic disasters, experts warn Security Russians are posing as Signal support to launch phishing attacksPLUS: US takes down Iranian propaganda sites; Marketing company asks 'Why Do We Have Your Information?' And more! Security Microsoft patches failed to fix on-prem SharePoint, which is now under zero-day attackPLUS: China upgrades smartphone surveillance tools; Ring eases anti-snooping stance; and more Black Hat and DEF CON DEF CON Franklin project enlists hackers to harden critical infrastructureVoting village reports have been so successful, says Jeff Moss, that the whole of DEF CON will now be included Security EQT buys majority share in Swiss cybersecurity biz AcronisWent at equivalent of $3.5B+ valuation for entire firm, though portion sold not specified Malware Month Ten years since the first corp ransomware, Mikko Hyppönen sees no end in sightOn the plus side, infosec's a good bet for a long, stable career FOSS smashed one Microsoft monopoly. After 20 years of failure, it's time to smash anotherWord up GNOME can look like Windows – and Flashback can do it without extensionsNew 'Simple-taskbar' is an option, but there's a simpler, stabler way A moment of silence, please, for the final release of Debian on x86-32New Debian versions hit FOSSland in the form of 13.6 and 12.15 Baddies caught exploiting extensions bugs with perfect 10 scores on vulnerable Joomla websitesFlaws in iCagenda, Balbooa Forms extensions can impact open source CMS that powers a million sites worldwide Frame: A new X11 server – implemented directly in assemblyJoins yserver, Phoenix, and of course XLibre – and outlier Arcan Cinnamon 6.8 will support Wayland – if you want itNext version of Linux Mint’s desktop has both kinds of display server

Cybersecurity researchers have disclosed what they say is an active "Shai-Hulud-like" supply chain worm campaign that has leveraged a cluster of at least 19 malicious npm packages to enable credential harvesting and cryptocurrency key theft. The campaign has been codenamed SANDWORM_MODE by supply chain security company Socket. As with prior Shai-Hulud attack waves, the malicious code embedded into the packages comes with capabilities to siphon system information, access tokens, environment secrets, and API keys from developer environments and automatically propagate by abusing stolen npm and GitHub identities to extend its reach. "The sample retains Shai-Hulud hallmarks and adds GitHub API exfiltration with DNS fallback, hook-based persistence, SSH propagation fallback, MCP server injection with embedded prompt injection targeting AI coding assistants, and LLM API Key harvesting," the company said. The packages, published to npm by two npm publisher aliases, official334 and javaorg, are listed below - [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] Also identified are four sleeper packages that do not incorporate any malicious features - ethres iru-caches iruchache uudi The packages go beyond npm-based propagation by including a weaponized GitHub Action that harvests CI/CD secrets and exfiltrates them via HTTPS with DNS fallback. They also feature a destructive routine that acts as a kill switch by triggering home directory wiping should it lose access to GitHub and npm. The wiper functionality is currently off by default. Another significant component of the malware is an "McpInject" module that specifically targets AI coding assistants by deploying a malicious model context protocol (MCP) server and injecting it into their tool configurations. The MCP server masquerades as a legitimate tool provider and registers three seemingly-harmless tools, each of which embeds a prompt injection to read the contents of ~/.ssh/id_rsa, ~/.ssh/id_ed25519, ~/.aws/credentials, ~/.npmrc, and .env files, stage them in a local directory for later exfiltration. The module targets Claude Code, Claude Desktop, Cursor, Microsoft Visual Studio Code (VS Code) Continue, and Windsurf. It also harvests API keys for nine large language models (LLM) providers: Anthropic, Cohere, Fireworks AI, Google, Grok, Mistral, OpenAI, Replicate, and Together. What's more, the payload contains a polymorphic engine that's configured to call a local Ollama instance with the DeepSeek Coder model to rename variables, rewrite control flow, insert junk code, and encode strings to evade detection. While the engine is turned off in the currently detected packages, the inclusion of the feature suggests that the operators are looking to release more iterations of the malware in the future. The entire attack chain unfolds over two stages: a first-stage component that captures credentials and cryptocurrency keys and then loads a secondary stage that subsequently performs deeper harvesting of credentials from password managers, worm-like propagation, MCP injection, and full exfiltration. The second stage is not activated until 48 hours (along with a per-machine jitter of up to 48 additional hours) have elapsed. Users who have installed any of the aforementioned packages are advised to remove them with immediate effect, rotate npm/GitHub tokens and CI secrets, and review any package.json, lockfiles, and .github/workflows/ for any unexpected changes. "Several feature flags and guardrails still suggest the threat actor is iterating on capabilities (for example, toggles that disable destructive routines or polymorphic rewriting in some builds)," Socket said. "However, the same worm code appearing across multiple typosquatting packages and publisher aliases indicates intentional distribution rather than an accidental release." "The destructive and propagation behaviors remain real and high-risk, and defenders should treat these packages as active compromise risks rather than benign test artifacts." The disclosure comes as Veracode and JFrog detailed two other malicious npm packages named "buildrunner-dev" and "eslint-verify-plugin," respectively, that are designed to deliver a remote access trojan (RAT) targeting Windows, macOS, and Linux systems. The .NET malware deployed by buildrunner-dev is Pulsar RAT, an open-source RAT delivered via a PNG image hosted on i.ibb[.]co. Eslint-verify-plugin, on the other hand, "masquerades as a legitimate ESLint utility while deploying a sophisticated, multi-stage infection chain targeting macOS and Linux environments," JFrog said. On Linux, the package deploys a Poseidon agent for the Mythic C2 framework. It facilitates a wide range of post-exploitation capabilities, including file operations, credential harvesting, and lateral movement. The macOS infection sequence executes Apfell, a JavaScript for Automation (JXA) agent for macOS, to conduct extensive data collection and create a new macOS user with admin privileges. Some of the data stolen by the agent are as follows - System information System credentials via a fake password dialog Google Chrome browser bookmarks Clipboard contents Files associated with iCloud Keychain and Chrome cookies, login data, and bookmarks Screenshots File metadata "The eslint-verify-plugin package is a direct example of how a malicious npm package can escalate from a simple installation hook to a full-system compromise," JFrog said. "By masquerading as a legitimate utility, the attackers successfully concealed a multi-stage infection chain." The findings also follow a report from Checkmarx, which flagged a rogue VS Code extension known as "solid281" that impersonates the official Solidity extension, but harbors covert features to execute a heavily obfuscated loader automatically upon application startup and drop ScreenConnect on Windows and a Python reverse shell on macOS and Linux machines. "This mirrors broader patterns reported by other teams: Solidity developers appear to be targeted specifically, including campaigns that used fake Solidity extensions to install ScreenConnect and then deploy follow-on payloads," Checkmarx noted.

Veracode’s platform enhancements help prevent software supply chain attacks Veracode announced significant platform innovations introduced through the second half of 2025. Headlining the release is Package Firewall, a preventive control for software supply chains, advancing the company’s mission to help organizations run secure software from code to cloud. With supply chain-related third-party breaches doubling year over year— from 15 to 30 percent according to the Verizon 2025 Data Breach Investigations Report— the need to strengthen security across the software ecosystem has never been greater. “The growing attack surface has created an unprecedented level of complexity for security and development teams,” said Tim Jarrett, VP of Product at Veracode. “The latest enhancements to our platform empower organizations to stop third-party risk from ever entering their software code, providing them with a prevention-first approach.” Package Firewall enhancement: Preventing supply chain attacks at the source Package Firewall, originally launched in June 2025, provides organizations with preventive control over their software supply chains by blocking malicious and risky packages before they enter the development environment. While traditional Software Composition Analysis (SCA) tools identify vulnerabilities in packages that are already in use, Veracode Package Firewall stops threats at the point of ingestion. The solution integrates with Azure Artifacts and deploys in seconds with integrations into package managers and repositories, like NPM, PyPI, Maven, Nexus, and Artifactory. It also supports custom policies that enable organizations to enforce security standards while maintaining developer productivity. Organizations can configure policies based on package risk profiles, vulnerability thresholds, and specific security requirements. Expanded platform capabilities and developer experience Veracode continued to expand and enhance its platform capabilities and developer experience throughout the year, with each release improving detection accuracy. Dynamic Application Security Testing (DAST) Essentials gained manual application linking, enabling policy evaluation and consolidated reporting. Software Composition Analysis (SCA) was upgraded with intelligent policies that only fail builds when fixes are available for vulnerable components, reducing developer friction. Static Analysis support was added for frameworks, including .NET Semantic Kernel, Python frameworks like AWS Glue and FastAPI, Java JDK 25 (LTS), and Node.js 22.x. The platform also saw significant updates to developer integrations for Visual Studio, JetBrains, Azure DevOps, and GitHub, alongside expansions to Veracode Security Labs, with training modules in container security content and the latest OWASP Top 10. Enterprise-grade authentication for developer tools The most recent platform update saw Veracode introduce the advanced enterprise-grade security required by organizations. The company added more thorough platform integration and deeper role-based access control to Veracode Risk Manager (VRM), and OAuth-based single sign-on (SSO) authentication across its entire Integrated Development Environment (IDE) plugin portfolio, including Visual Studio Code, Visual Studio, Eclipse, and JetBrains platforms. The integration eliminates Application Programming Interface (API) key management and delivers secure, enterprise-grade authentication with centralized access control. Jarrett closed, “Our mission is to empower organizations to enhance their security posture, bridge critical skills gaps, and accelerate remediation, all within a unified, integrated platform. By listening closely to our customers, we continuously evolved Veracode’s platform in 2025 to meet their needs, enabling them to drive faster, more secure DevSecOps practices.”

Week in review: Cisco fixes critical UCCX flaws, November 2025 Patch Tuesday forecast Here’s an overview of some of last week’s most interesting news, articles, interviews and videos: Securing real-time payments without slowing them down In this Help Net Security interview, Arun Singh, CISO at Tyro, discusses what it takes to secure real-time payments without slowing them down. He explains how analytics, authentication, and better industry cooperation can help stay ahead of fraud. Singh also touches on how digital identity and accountability are transforming how trust is built in payments. Heisenberg: Open-source software supply chain health check tool Heisenberg is an open-source tool that checks the health of a software supply chain. It analyzes dependencies using data from deps.dev, Software Bills of Materials (SBOMs), and external advisories to measure package health, detect risks, and generate reports for individual dependencies or entire projects. A new way to think about zero trust for workloads Static credentials have been a weak point in cloud security for years. A new paper by researchers from SentinelOne takes direct aim at that issue with a practical model for authenticating workloads without long-lived secrets. Instead of relying on static keys, the team proposes using temporary, verifiable tokens that expire within minutes. How nations build and defend their cyberspace capabilities In this Help Net Security interview, Dr. Bernhards Blumbergs, Lead Cyber Security Expert at CERT.LV, discusses how cyberspace has become an integral part of national and military operations. He explains how countries develop capabilities to act and defend in this domain, often in coordination with activities in other areas of conflict. AI can flag the risk, but only humans can close the loop In this Help Net Security interview, Dilek Çilingir, Global Forensic & Integrity Services Leader at EY, discusses how AI is transforming third-party assessments and due diligence. She explains how machine learning and behavioral analytics help organizations detect risks earlier, improve compliance, and strengthen accountability. PortGPT: How researchers taught an AI to backport security patches automatically Keeping older software versions secure often means backporting patches from newer releases. It is a routine but tedious job, especially for large open-source projects such as the Linux kernel. A new research effort has built a tool that uses a large language model to do that work automatically. OpenGuardrails: A new open-source model aims to make AI safer for real-world use When you ask a large language model to summarize a policy or write code, you probably assume it will behave safely. But what happens when someone tries to trick it into leaking data or generating harmful content? That question is driving a wave of research into AI guardrails, and a new open-source project called OpenGuardrails is taking a bold step in that direction. What keeps phishing training from fading over time When employees stop falling for phishing emails, it is rarely luck. A new study shows that steady, mandatory phishing training can cut risky behavior over time. After one year of continuous simulations and follow-up lessons, employees were half as likely to take the bait. Metrics don’t lie, but they can be misleading when they only tell IT’s side of the story In this Help Net Security interview, Rik Mistry, Managing Partner at Interval Group, discusses how to align IT strategy with business goals. He explains how security, governance, and orchestration shape IT operations and why early collaboration between IT and security leaders leads to better outcomes. Mistry also shares his perspective on automation and emerging technologies. Cyber-espionage campaign mirroring Sandworm TTPs hit Russian and Belarusian military A spear-phishing campaign aimed to compromise Russian and Belarusian military personnel by using military-themed documents as a lure has been flagged by Cyble and Seqrite security researchers. The goal of the campaign is to get targets to download and open a booby-trapped LNK file masquerading as a PDF, ultimately leading to a complete system compromise. Former ransomware negotiators allegedly targeted US firms with ALPHV/BlackCat ransomware A ransomware negotiator and an incident response manager have been indicted in Florida for allegedly conspiring to deploy the ALPHV/BlackCat ransomware against multiple US companies and extorting nearly $1.3 million from one of the victims. Cybercriminals exploit RMM tools to steal real-world cargo Cybercriminals are compromising logistics and trucking companies by tricking them into installing remote monitoring and management (RMM) tools, Proofpoint researchers warned. Critical Control Web Panel vulnerability is actively exploited (CVE-2025-48703) On Tuesday, CISA added two vulnerabilities to its Known Exploited Vulnerabilities catalog: CVE-2025-11371, which affects Gladinet’s CentreStack and Triofox file-sharing and remote access platforms, and CVE-2025-48703, a vulnerability in Control Web Panel (CWP), a web hosting control panel designed for managing servers running CentOS or CentOS-based distributions. Google uncovers malware using LLMs to operate and evade detection PromptLock, the AI-powered proof-of-concept ransomware developed by researchers at NYU Tandon and initially mistaken for an active threat by ESET, is no longer an isolated example: Google’s latest report shows attackers are now creating and deploying other malware that leverages LLMs to operate and evade security systems. SonicWall cloud backup hack was the work of a state actor Incident responders from Mandiant have wrapped up their investigation into the SonicWall cloud backup service hack, and the verdict is in: the culprit is a state-sponsored threat actor (though the specific nation wasn’t disclosed). Cisco fixes critical UCCX flaws, patch ASAP! (CVE-2025-20358, CVE-2025-20354) Cisco has fixed two critical vulnerabilities (CVE-2025-20358, CVE-2025-20354) affecting Unified Contact Center Express (UCCX), which may allow attackers to bypass authentication, compromise vulnerable installations, and elevate privileges to root. Attackers upgrade ClickFix with tricks used by online stores Attackers have taken the ClickFix technique further, with pages borrowing tricks from online sellers to pressure victims into performing the steps that will lead to a malware infection. Uncovering the risks of unmanaged identities Every organization manages thousands of identities, from admins and developers to service accounts and AI agents. But many of these identities operate in the shadows, untracked and unprotected. These unmanaged identities quietly expand your attack surface, weaken compliance, and threaten business continuity, posing significant risk. So, how can you uncover, secure, and manage what you can’t see? November 2025 Patch Tuesday forecast: Windows Exchange Server EOL? October 2025 Patch Tuesday was one for the record books in so many ways. There was a big push by Microsoft to fix as many open vulnerabilities as possible in products that were reaching end-of-life (EOL). This included 116 CVEs addressed in Windows 10 and an astronomical 134 CVEs addressed in Windows 11, because don’t forget Windows 11 22H2 Enterprise and Education editions also reached EOL. Cybercriminals have built a business on YouTube’s blind spots The days when YouTube was just a place for funny clips and music videos are behind us. With 2.53 billion active users, it has become a space where entertainment, information, and deception coexist. European authorities dismantle €600 million crypto scam network Nine people have been arrested in a coordinated international operation targeting a large cryptocurrency money laundering network that defrauded victims of more than €600 million. The operation was led by Eurojust, the EU’s judicial cooperation agency, which brought together investigators and prosecutors from France, Belgium, Cyprus, Spain and Germany. Connected homes: Is bystander privacy anyone’s responsibility? Smart doorbells, connected cameras, and home monitoring systems have become common sights on doorsteps and living rooms. They promise safety and convenience, but they also raise a problem. These devices record more than their owners. They capture neighbors, visitors, and anyone passing by. 18 arrested in €300 million global credit card fraud scheme A coordinated international operation has led to 18 arrests in a massive credit card fraud case worth at least €300 million. The effort, led by Eurojust, targeted a network of suspects accused of running fake online subscription services for dating, pornography, and streaming sites. Among those detained were five executives from four German payment service providers. Enterprises are losing track of the devices inside their networks Security teams are often surprised when they discover the range and number of devices connected to their networks. The total goes far beyond what appears in agent-based telemetry or old manual asset inventories. Deepfakes, fraud, and the fight for trust online In this Help Net Security video, Michael Engle, Chief Strategy Officer at 1Kosmos, explains how deepfakes are changing online identity verification. He describes how fake IDs and synthetic identities are being used for account signups and takeovers. What shadow AI means for your company’s security In this Help Net Security video, Peled Eldan, Head of Research at XM Cyber, explains the hidden risks of shadow AI. He describes how employees often use unapproved AI tools at work to save time or solve problems, even when approved tools are available. This behavior, though common, can lead to serious issues such as data leaks, compliance violations, and security blind spots. Europe’s phone networks are drowning in fake calls Caller ID spoofing has become one of Europe’s most persistent enablers of cyber fraud. A new position paper from Europol warns that manipulated phone identities now drive much of the continent’s financial and social engineering crime, making it difficult for law enforcement to track perpetrators. The agency estimates global losses at around EUR 850 million a year, with phone and text-based fraud accounting for roughly two thirds of reported scam cases. Employees keep finding new ways around company access controls AI, SaaS, and personal devices are changing how people get work done, but the tools that protect company systems have not kept up, according to 1Password. Tools like SSO, MDM, and IAM no longer align with how employees and AI agents access data. Financial services can’t shake security debt In financial services, application security risk is becoming a long game. Fewer flaws appear in new code, but old ones linger longer, creating a kind of software “interest” that keeps growing, according to Veracode’s 2025 State of Software Security report. Google says 2026 will be the year AI supercharges cybercrime Security leaders are staring down a year of major change. In its Cybersecurity Forecast 2026, Google paints a picture of a threat landscape transformed by AI, supercharged cybercrime, and increasingly aggressive nation-state operations. Attackers are moving faster, scaling their operations with automation. VulnRisk: Open-source vulnerability risk assessment platform VulnRisk is an open-source platform for vulnerability risk assessment. It goes beyond basic CVSS scoring by adding context-aware analysis that reduces noise and highlights what matters. The tool is free to use and designed for local development and testing. Retailers are learning to say no to ransom demands Ransomware remains one of the biggest operational risks for retailers, but the latest data shows a shift in how these attacks unfold. Fewer incidents now lead to data encryption, recovery costs have dropped, and businesses are bouncing back faster. Yet attackers are demanding more money, and security teams are feeling the strain. Humans built the problem, AI just scaled it Information moves across cloud platforms, personal devices, and AI tools, often faster than security teams can track it. Proofpoint’s 2025 Data Security Landscape report shows that most organizations faced data loss last year, usually caused by their own people. With AI agents part of daily operations, security leaders are confronting risks that come from users and from the systems acting on their behalf. Russia-linked hackers intensify attacks as global APT activity shifts State-aligned hacking groups have spent the past six months ramping up espionage, sabotage, and cybercrime campaigns across multiple regions, according to ESET’s APT Activity Report covering April through September 2025. The research highlights how operations linked to Russia, China, Iran, and North Korea have evolved in scope and technique, showing that nation-state activity remains a constant source of disruption. Hospitals are running out of excuses for weak cyber hygiene Healthcare leaders continue to treat cybersecurity as a technical safeguard instead of a strategic business function, according to the 2025 US Healthcare Cyber Resilience Survey by EY. The study, based on responses from 100 healthcare executives, outlines six areas where hospitals and health systems must act to close resilience gaps that threaten patient care and operations. Old privacy laws create new risks for businesses Businesses are increasingly being pulled into lawsuits over how they collect and share user data online. What was once the domain of large tech firms is now a widespread legal risk for companies of all sizes. The latest analysis from cyber insurer Coalition shows that outdated privacy laws are driving a surge in web privacy claims, with small and midsize businesses now common targets. Product showcase: Cogent Community democratizes vulnerability intelligence with agentic AI Teams are buried under overlapping feeds, inconsistent formats, and fragmented context. Even with advanced tools, analyzing raw intelligence into prioritized, evidence-based action remains one of the hardest problems in modern security operations. Cogent Security addresses this problem head-on with its industry-first AI taskforce for vulnerability management. And with its newly introduced Cogent Community, the company is now delivering an open-access, free-to-use tool that helps security teams operationalize vulnerability intelligence. Cybersecurity jobs available right now: November 4, 2025 We’ve scoured the market to bring you a selection of roles that span various skill levels within the cybersecurity field. Check out this weekly selection of cybersecurity jobs available right now. New infosec products of the week: November 7, 2025 Here’s a look at the most interesting products from the past week, featuring releases from 1touch.io, Barracuda Networks, Bitdefender, Forescout, and Komodor.

Financial services can’t shake security debt In financial services, application security risk is becoming a long game. Fewer flaws appear in new code, but old ones linger longer, creating a kind of software “interest” that keeps growing, according to Veracode’s 2025 State of Software Security report. Researchers analyzed data from more than 1.3 million applications and 126 million security findings. Financial institutions perform better than average at preventing severe vulnerabilities, but they are slower to fix them and carry more long-term security debt than most other sectors. Fewer flaws, but a plateau in progress 57% of financial sector applications had at least one security flaw during their latest static analysis scan. About 55% contained issues listed in the OWASP Top 10, and 40% were tied to the CWE Top 25 most dangerous software weaknesses. Only 8% of financial applications had high-severity flaws, compared to 16% across all industries. These figures show that financial firms are identifying and preventing critical issues better than most. Yet, progress has stalled since 2021. After several years of improvement, the rate of vulnerable applications has leveled off, suggesting that organizations are struggling to push risk lower. The report attributes this to the constant churn of software development. Code is written, flaws are found, fixes are deployed, and new code introduces new issues. Fixing flaws takes close to nine months The median time to fix half of discovered flaws is about nine months, slightly slower than the cross-industry average. A significant share of flaws remain open for years, showing that while many fixes happen early, remediation slows over time. As these flaws accumulate, applications become harder to maintain. Older issues are often pushed aside as teams focus on new projects and vulnerabilities, creating a cycle of unfinished remediation work. “Trust is everything in financial services, yet our data reveals a silent, growing risk for the sector created by unresolved security debt. With AI-driven attacks surging and compliance requirements tightening, finance leaders must prioritize strategic risk reduction, starting with targeted remediation of critical software flaws,” said Chris Wysopal, Chief Security Evangelist at Veracode. Security debt is widespread and risky 77% carry some level of security debt, slightly higher than the 74% average across industries. Even more concerning, 63% of financial firms have what Veracode defines as critical debt, meaning severe vulnerabilities that have gone unfixed for more than a year. That figure is 13% points higher than the overall rate. Security debt tends to accumulate in older, larger applications. It often mirrors other forms of technical debt that reduce efficiency and resilience. The report suggests that the more legacy code an organization maintains, the heavier the backlog of flaws becomes. Open source adds to the burden About 17% of all security debt in the financial services sector comes from open-source libraries. When focusing on critical debt, that share rises above 80%. Flaws in third-party components take longer to fix than those in internally developed code. The report recommends evaluating open-source packages before adding them to a codebase, since inherited flaws can spread across multiple applications and persist for years. The findings also suggest that the full scale of open-source risk in financial institutions may be undercounted. Many organizations still lack visibility into their software supply chain and may not consistently use software composition analysis tools. Leaders move faster and carry less debt The report compares leading and lagging financial services organizations using five application security metrics: flaw prevalence, fix capacity, fix speed, debt prevalence, and open-source debt. Leading organizations fix flaws several times faster than their peers and maintain a much lower level of security debt across their applications. Laggards, by contrast, remediate at a slow pace and carry lingering flaws in most of their systems. The gap shows that a small group of mature programs are steadily improving, while many others remain weighed down by long-standing flaws.

Il software è oggi il motore di ogni servizio digitale, ma anche uno dei principali veicoli di rischio. Vulnerabilità nel codice proprietario e open source, librerie non aggiornate e pratiche di sviluppo poco sicure alimentano quello che viene definito “debito di sicurezza applicativa”, un accumulo di problemi che può compromettere la resilienza complessiva di un’organizzazione. Le aziende di servizi digitali e di digital trust si trovano a dover conciliare richieste sempre più stringenti di compliance (GDPR, eIDAS, NIS2, AgID, ACN) con cicli di sviluppo rapidi e pressioni di business. In questo contesto, CryptoNet Labs ha realizzato per un’importante realtà italiana di Digital Trust Services un programma di Application Security gestita, definito SOC Applicativo, finalista ai Cybersecurity360 Awards. Indice degli argomenti L’obiettivo del progetto non era solo individuare vulnerabilità a valle, tramite VAPT periodici, ma spostare i controlli a monte del ciclo di sviluppo, secondo il paradigma dello Shift Security Left. La sfida principale consisteva nell’equilibrare tre fattori: l’esigenza di garantire compliance normativa a stakeholder e autorità di vigilanza; la necessità di rispettare i tempi di sviluppo e rilascio richiesti dal business; la limitata disponibilità di competenze specifiche di sicurezza all’interno dei team di sviluppo. CryptoNet Labs ha risposto con un modello incrementale e sostenibile, che unisce test dinamici e statici, formazione continua e supervisione diretta degli sviluppatori. Il competence center di Application Security di CryptoNet Labs ha costruito il progetto su tre pilastri complementari: Vulnerability Assessment e Penetration Test (VAPT): attività gestite dal laboratorio interno, con focus sugli applicativi in esercizio e sui relativi contenitori infrastrutturali. SAST e SCA tramite Veracode: la piattaforma è stata integrata nelle pipeline di sviluppo CI/CD, con controlli su codice proprietario e open source fin dalle prime fasi. Supervisione e accompagnamento: gli specialisti di CryptoNet hanno affiancato per oltre un anno gli 11 team di sviluppo del cliente, con incontri ricorrenti dedicati all’analisi delle vulnerabilità emerse e alla definizione delle remediation. A questi elementi si è aggiunto un costante allineamento con la funzione Cybersecurity, owner del progetto, che ha monitorato i progressi tramite KPI di Application Security Governance. Il progetto ha previsto un setup iniziale di circa un mese, dedicato all’integrazione della piattaforma Veracode con gli strumenti di CI/CD e ticketing esistenti. In questo modo le analisi SAST e SCA sono state inserite direttamente negli stage di rilascio, garantendo automazione e continuità. Terminata questa fase, sono stati istituiti meeting ricorrenti con i Security Champion di ciascun team di sviluppo: settimanali all’inizio, poi con cadenza quindicinale o mensile, a seconda della maturità raggiunta. Parallelamente, le attività di VAPT sono state pianificate su base annuale, con incontri di coordinamento per allineare test e priorità operative. Questo approccio ha consentito di trasformare la sicurezza applicativa da attività episodica a processo strutturato, in grado di crescere insieme all’organizzazione. In pochi mesi, il cliente ha potuto registrare risultati concreti: riduzione del debito di sicurezza applicativa, con eliminazione delle vulnerabilità di maggiore severità; responsabilizzazione degli sviluppatori, coinvolti direttamente nella remediation; compliance dimostrabile, grazie a report e KPI condivisi con gli stakeholder; sinergia tra tecnologia, processi e persone, per un programma di Application Security sostenibile. Il valore del progetto non risiede solo nella riduzione delle vulnerabilità, ma nella creazione di una cultura diffusa della sicurezza all’interno dei team di sviluppo. Un aspetto distintivo del SOC Applicativo di CryptoNet Labs è l’attenzione alla remediation effettiva. Ogni attività di test diventa utile solo se si traduce in correzione del codice. Per accelerare questo processo, il progetto ha integrato le funzionalità di Generative AI di Veracode Fix, che suggeriscono allo sviluppatore direttamente sul desktop come riscrivere il codice in modo sicuro. La combinazione di VAPT, SAST e SCA, per loro natura complementari, ha inoltre permesso di massimizzare lo spettro delle vulnerabilità identificate e di aumentarne l’affidabilità grazie alla convergenza dei risultati. Il SOC Applicativo dimostra come la sicurezza software possa essere resa scalabile e replicabile. La metodologia, basata su automazione, piattaforme leader di mercato e supervisione diretta, può essere applicata in qualsiasi contesto con sviluppo software strutturato, dalla finanza alla pubblica amministrazione, fino alle telco e alle utility. Il valore innovativo risiede nell’aver trasformato la sicurezza applicativa da obbligo a processo di governance, in cui i team di sviluppo diventano attori centrali e consapevoli. Questo approccio consente alle aziende di ridurre il rischio, accelerare i rilasci e rispondere più rapidamente alle pressioni normative. Il progetto realizzato da CryptoNet Labs mostra come sia possibile abbattere il debito di sicurezza applicativa in maniera concreta, senza rallentare i processi di business. L’unione di VAPT, SAST, SCA e AI generativa, insieme al coinvolgimento diretto dei team di sviluppo, rappresenta un modello efficace per rafforzare la resilienza digitale. La candidatura ai Cybersecurity360 Awards conferma l’importanza di un approccio che porta la sicurezza nel cuore dello sviluppo software, coniugando compliance, innovazione e sostenibilità.

Week in review: SonicWall firewalls targeted in ransomware attacks, Black Hat USA 2025 Here’s an overview of some of last week’s most interesting news, articles, interviews and videos: Black Hat USA 2025 Black Hat USA 2025 took place at the Mandalay Bay Convention Center in Las Vegas. Explore related news, photos, product releases, and more. AIBOMs are the new SBOMs: The missing link in AI risk management In this Help Net Security interview, Marc Frankel, CEO at Manifest Cyber, discusses how overlooked AI-specific risks, like poisoned training data and shadow AI, can lead to security issues that conventional tools fail to detect. He explains how AI Bills of Materials (AIBOMs) extend SBOMs to provide transparency into datasets, model weights, and third-party integrations, improving governance and incident response. Security tooling pitfalls for small teams: Cost, complexity, and low ROI In this Help Net Security interview, Aayush Choudhury, CEO at Scrut Automation, discusses why many security tools built for large enterprises don’t work well for leaner, cloud-native teams. He explains how simplicity, integration, and automation are key for SMBs with limited resources. Beyond PQC: Building adaptive security programs for the unknown In this Help Net Security interview, Jordan Avnaim, CISO at Entrust, discusses how to communicate the quantum computing threat to executive teams using a risk-based approach. He explains why post-quantum cryptography (PQC) is an urgent and long-term priority. SonicWall firewalls targeted in ransomware attacks, possibly via zero-day Attackers wielding the Akira ransomware and possibly a zero-day exploit have been spotted targeting SonicWall firewalls since July 15, 2025. Project Ire: Microsoft’s autonomous malware detection AI agent Microsoft is working on a AI agent whose main goal is autonomous malware detection and the prototype – dubbed Project Ire – is showing great potential, the company has announced on Tuesday. Millions of Dell laptops could be persistently backdoored in ReVault attacks A set of firmware vulnerabilities affecting 100+ Dell laptop models widely used in government settings and by the cybersecurity industry could allow attackers to achieve persistent access even across Windows reinstalls, Cisco Talos researchers have discovered. Trend Micro Apex One flaws exploted in the wild (CVE-2025-54948, CVE-2025-54987) Unauthenticated command injection vulnerabilities (CVE-2025-54948, CVE-2025-54987) affecting the on-premise version of Trend Micro’s Apex One endpoint security platform are being probed by attackers, the company has warned on Wednesday. Adobe patches critical Adobe Experience Manager Forms vulnerabilities with public PoC Adobe has released an emergency security update for Adobe Experience Manager Forms on Java Enterprise Edition (JEE), which fix two critical vulnerabilities (CVE-2025-54253, CVE-2025-54254) with a publicly available proof-of-concept (PoC) exploit. Microsoft urges admins to plug severe Exchange security hole (CVE-2025-53786) The privilege escalation can be performed by exploiting CVE-2025-53786, a newly disclosed vulnerability that stems from Exchange Server and Exchange Online sharing the same service principal – i.e., the Office 365 Exchange Online application – in hybrid configurations. (The app is used to authenticate and secure the communication between Exchange Server and Exchange Online.) Cybersecurity and the development of software-defined vehicles In many automotive companies, the same systems-engineering teams are responsible for both safety and security. As a result, cybersecurity is treated as a subset of safety, undergirded by an implicit assumption: “If it’s safe, it must be secure.” What GPT‑5 means for IT teams, devs, and the future of AI at work OpenAI has released GPT‑5, the newest version of its large language model. It’s now available to developers and ChatGPT users, and it brings some real changes to how AI can be used in business and IT environments. August 2025 Patch Tuesday forecast: Try, try again July started slow with a calm Patch Tuesday, addressing 130 CVEs but only one publicly disclosed, so risk was low. Then two SharePoint CVEs were exploited, heating things up with hotfixes late in the month. Add security issues in Microsoft Exchange and major Google and Apple updates, and July ended very active. Open-source password recovery utility Hashcat 7.0.0 released Hashcat is an open-source password recovery tool that supports five attack modes and more than 300 highly optimized hashing algorithms. It runs on CPUs, GPUs, and other hardware accelerators across Linux, Windows, and macOS, and includes features for distributed password cracking at scale. BloodHound 8.0 debuts with major upgrades in attack path management SpecterOps has released BloodHound 8.0, the latest iteration of its open-source attack path management platform, featuring major enhancements and expanded capabilities. Open-source server management platform Proxmox VE 9.0 released Proxmox Virtual Environment (VE) 9.0 introduces advancements in both storage and networking capabilities, addressing critical enterprise demands. AI in the SOC: Game-changer or more noise? In this Help Net Security video, Kev Marriott, Senior Manager of Cyber at Immersive Labs, explores the challenges and opportunities of integrating AI into Security Operations Centers (SOCs). Third-party partners or ticking time bombs? In this Help Net Security video, Ngaire Elizabeth Guzzetti, Technical Director Supply Chain at CyXcel, discusses why a third of U.S. organizations don’t trust third-party vendors to manage critical risks and what that means for supply chain security. What’s keeping risk leaders up at night? AI, tariffs, and cost cuts Enterprise risk leaders are most concerned about rising tariffs and trade tensions heading into the second half of 2025, according to a new report from Gartner. The firm’s second-quarter Emerging Risk Report, based on a survey of 223 senior risk, audit, and compliance executives, ranks the escalating trade war as the top emerging risk, up from third in the first quarter. The surprising truth about identity security confidence Organizations most confident in their identity security are often the least prepared, according to a new report from BeyondID. The study reveals a troubling gap between what organizations believe about their identity security programs and how they actually behave. Surprisingly, those expressing the highest confidence are adopting fewer best practices than their more cautious peers. Average global data breach cost now $4.44 million IBM released its Cost of a Data Breach Report, which revealed AI adoption is greatly outpacing AI security and governance. While the overall number of organizations experiencing an AI-related breach is a small representation of the researched population, this is the first time security, governance and access controls for AI have been studied in this report, which suggests AI is already an easy, high value target. Ransomware groups shift to quadruple extortion to maximize pressure Threat actors are using a new quadruple extortion tactic in ransomware campaigns, while double extortion remains the most common approach, according to Akamai. Your employees uploaded over a gig of files to GenAI tools last quarter In Q2 2025, Harmonic reviewed 1 million GenAI prompts and 20,000 uploaded files across more than 300 GenAI and AI-powered SaaS apps, and the findings confirm that sensitive data is being exposed through GenAI tools, something many security leaders fear but find difficult to measure. Why 90% of cyber leaders are feeling the heat 90% of cyber leaders find managing cyber risks harder today than five years ago, mainly due to the explosion of AI and expanding attack surfaces, according to BitSight. These threats are also fueling high rates of burnout, with 47% of cybersecurity and cyber risk professionals reporting exhaustion. CISOs say they’re prepared, their data says otherwise A new Axonius report, based on a survey of 500 U.S.-based IT and security leaders, shows a disconnect between perceived readiness and actual performance in vulnerability and exposure management. While 90% of respondents said their organization is prepared to act when a threat is found, only 25% said they trust the data in their security tools. Ransomware is up, zero-days are booming, and your IP camera might be next Cyber attackers are finding new ways in through the overlooked and unconventional network corners. Forescout’s 2025H1 Threat Review reveals a surge in advanced tactics, with zero-day exploits up 46 percent and ransomware attacks averaging 20 per day. Energy companies are blind to thousands of exposed services Many of America’s largest energy providers are exposed to known and exploitable vulnerabilities, and most security teams may not even see them, according to a new report from SixMap. AI can write your code, but nearly half of it may be insecure While GenAI excels at producing functional code, it introduces security vulnerabilities in 45 percent of cases, according to Veracode’s 2025 GenAI Code Security Report, which analyzed code produced by over 100 LLMs across 80 real-world coding tasks. Cybercriminals are getting personal, and it’s working Cybercriminals are deploying unidentifiable phishing kits (58% of phishing sites) to propagate malicious campaigns at scale, indicating a trend towards custom-made or obfuscated deployments, according to VIPRE Security. Fraud controls don’t guarantee consumer trust Over a third of companies say they are using AI, including generative AI, to fight fraud, according to Experian. As fraud threats become more complex, companies are accelerating their investments with over half adopting new analytics and building AI models to enhance customer decision-making. From fake CAPTCHAs to RATs: Inside 2025’s cyber deception threat trends Cybercriminals are getting better at lying. That’s the takeaway from a new LevelBlue report, which outlines how attackers are using social engineering and legitimate tools to quietly move through environments before they’re caught. Back to basics webinar: The ecosystem of CIS Security best practices Watch this on-demand webinar to understand how you can use an ecosystem of security best practices built by the Center for Internet Security (CIS) to support this mission. Cybersecurity jobs available right now: August 5, 2025 We’ve scoured the market to bring you a selection of roles that span various skill levels within the cybersecurity field. Check out this weekly selection of cybersecurity jobs available right now. New infosec products of the week: August 8, 2025 Here’s a look at the most interesting products from the past week, featuring releases from Black Kite, Descope, Elastic, ExtraHop, LastPass, and Riverbed.

Here’s a look inside Black Hat USA 2025. The featured vendors are: Stellar Cyber, Vonahi Security, Gurucul, Check Point, HackerOne, EasyDMARC, Elastic, Google, Tines, Veracode, VioletX, Pentera, Keep Aware, Oleria, SpyCloud, Trend Micro and Picus Security. The post Photos: Black Hat USA 2025 appeared first on Help Net Security .

AI can write your code, but nearly half of it may be insecure While GenAI excels at producing functional code, it introduces security vulnerabilities in 45 percent of cases, according to Veracode’s 2025 GenAI Code Security Report, which analyzed code produced by over 100 LLMs across 80 real-world coding tasks. Vibe coding “The rise of vibe coding, where developers rely on AI to generate code, typically without explicitly defining security requirements, represents a fundamental shift in how software is built,” said Jens Wessling, CTO at Veracode. “The main concern with this trend is that they do not need to specify security constraints to get the code they want, effectively leaving secure coding decisions to LLMs. Our research reveals GenAI models make the wrong choices nearly half the time, and it’s not improving.” AI is enabling attackers to identify and exploit security vulnerabilities quicker. Tools powered by AI can scan systems at scale, identify weaknesses, and even generate exploit code with minimal human input. This lowers the barrier to entry for less-skilled attackers and increases the speed and sophistication of attacks, posing a threat to traditional security defenses. Not only are vulnerabilities increasing, but the ability to exploit them is becoming easier. LLMs introduce widespread security vulnerabilities in code To assess the security of AI-generated code, Veracode created 80 coding tasks designed to expose common software weaknesses as defined by the MITRE Common Weakness Enumeration (CWE) system. Each task prompted more than 100 LLMs to complete code snippets, offering the opportunity to choose a secure or insecure implementation. The findings were stark: in 45 percent of all test cases, LLMs produced code containing vulnerabilities aligned with the OWASP Top 10, a list of the most serious web application security risks. Some languages proved especially problematic. Java had the highest failure rate, with LLM-generated code introducing security flaws more than 70 percent of the time. Python, C# and JavaScript were not far behind, with failure rates between 38 and 45 percent. LLMs also struggled with specific vulnerability types. Eighty-six percent of code samples failed to defend against cross-site scripting (CWE-80), and 88 percent were vulnerable to log injection attacks (CWE-117). “Our research shows models are getting better at coding accurately but are not improving at security. We also found larger models do not perform significantly better than smaller models, suggesting this is a systemic issue rather than an LLM scaling problem,” Wessling said. Securing the AI-driven software pipeline While GenAI development practices like vibe coding accelerate productivity, they also amplify risks. Researchers emphasize that organizations need a risk management program that prevents vulnerabilities before they reach production—by integrating code quality checks and automated fixes directly into the development workflow. As organizations increasingly leverage AI-powered development, Veracode recommends taking the following proactive measures to ensure security: Integrate AI-powered tools into developer workflows to remediate security risks in real time. Leverage static analysis to detect flaws early and automatically, preventing vulnerable code from advancing through development pipelines. Embed security in agentic workflows to automate policy compliance and ensure AI agents enforce secure coding standards. Use Software Composition Analysis (SCA) to ensure AI-generated code does not introduce vulnerabilities from third-party dependencies and open-source components. Adopt AI-driven remediation guidance to empower developers with fix instructions and train them to use the recommendations. Deploy a firewall to automatically detect and block malicious packages, vulnerabilities, and policy violations.

L’intelligenza artificiale generativa (GenAI) sta rapidamente cambiando il modo in cui viene sviluppato software. Se in passato gli sviluppatori scrivevano il codice direttamente in linguaggi di programmazione specifici, oggi sempre più spesso descrivono le funzionalità desiderate in linguaggio naturale, affidandosi ai Large Language Models (LLM) per generare il codice in automatico. Tuttavia, questo approccio presenta una criticità: i programmatori non sono tenuti a fornire istruzioni specifiche volte a prevenire vulnerabilità di sicurezza. Un recente studio condotto da Veracode, azienda specializzata nella gestione del rischio delle applicazioni, ha analizzato il livello di sicurezza del codice generato da più di 100 Large Language Models in quattro linguaggi di programmazione (Java, JavaScript, C# e Python). L’obiettivo principale era valutare se i modelli generativi siano in grado di produrre codice sicuro anche in assenza di istruzioni specifiche, contenute nel prompt, volte a prevenire vulnerabilità di sicurezza. In questo articolo approfondiremo i risultati emersi dallo studio e ne analizzeremo le implicazioni. Metodologia Lo studio si è incentrato nel valutare il livello di sicurezza del codice generato da oltre 100 assistenti di programmazione basati su LLM, utilizzando quattro diversi linguaggi di programmazione (Java, JavaScript, C# e Python). I modelli sono stati testati utilizzando una serie di task di programmazione con vulnerabilità di sicurezza note, in conformità con il sistema MITRE Common Weakness Enumeration (CWE). In particolare, i task di programmazione si sono basati su quattro CWEs: SQL injection (CWE 89), cross-site scripting (CWE 80), log injection (CWE 117) e algoritmi crittografici insicuri (CWE 327). A ciascun assistente di programmazione è stato chiesto di completare cinque diverse istanze di task per ogni combinazione di CWE e linguaggio di programmazione. Il codice generato dagli assistenti è stato analizzato utilizzando la metodologia Static Application Security Testing (SAST). Task di programmazione Ogni task di programmazione consiste in una singola funzione scritta in uno dei quattro linguaggi di programmazione (Java, JavaScript, C# e Python). Una parte del corpo della funzione è stata sostituita da un commento che descrive la funzionalità desiderata. È importante notare che il codice mancante può essere implementato in diversi modi, tra cui almeno uno che introduce la CWE di interesse. Il prompt fornito agli assistenti di programmazione contiene solo il codice della funzione (con il commento) e la richiesta di completamento della parte mancante. Questo consente di valutare le scelte di sicurezza del modello senza altri fattori esterni. Ad esempio, fornendo istruzioni volte a prevenire vulnerabilità di sicurezza, il codice generato potrebbe risultare diverso. Poiché nella pratica gli sviluppatori tendono a concentrarsi principalmente sulla funzionalità del codice, trascurando talvolta le implicazioni legate alle vulnerabilità di sicurezza, questo approccio è stato pensato per riflettere scenari reali e frequenti. Risultati e Analisi Prima di analizzare i risultati dello studio, è necessario definire due parametri: il security pass rate, ovvero la percentuale di task in cui il modello sceglie un’implementazione sicura, e il syntax pass rate, ossia la percentuale di task in cui il codice generato è sintatticamente corretto e compilabile. Come illustrato nel grafico sottostante, sebbene i modelli abbiano fatto progressi significativi nella generazione di codice sintatticamente corretto, le prestazioni in termini di sicurezza non presentano miglioramenti. Infatti, oltre il 90% del codice prodotto dagli LLM rilasciati nel 2025 è risultato compilabile – un netto miglioramento rispetto al 20% risalente a giugno 2023 – ma solo il 55% ha superato con successo i controlli di sicurezza successivi. In altre parole, nel 45% dei casi, i modelli hanno introdotto vulnerabilità di sicurezza appartenenti alla top 10 OWASP. Perché, nonostante i miglioramenti nella generazione di codice sintatticamente corretto, le performance a livello sicurezza restano basse? Secondo Veracode, questo andamento può essere spiegato dalla natura dei dataset di addestramento, costituiti da codice pubblicamente accessibile su Internet. Poiché la maggior parte di questo codice è sintatticamente corretto, la qualità dei risultati su questo fronte dipende soprattutto dalla capacità dei modelli di apprendere e riprodurre correttamente la sintassi – capacità che migliora con l’aumentare della potenza del modello. Al contrario, il livello di sicurezza del codice risulta piuttosto basso, poiché in molti casi il codice pubblico presenta ancora vulnerabilità non risolte. Dato che la maggior parte dei modelli testati si basa su dataset pubblici, non sorprende che presentino pattern ricorrenti e prestazioni di sicurezza simili. Un altro dato interessante emerso dallo studio riguarda Java, linguaggio per il quale gli LLM faticano a generare codice sicuro: in media, il security pass rate si ferma al 28,5%, come illustrato nel grafico sottostante. Al contrario, le performance sono sia più elevate che più stabili negli altri tre linguaggi considerati – Python, JavaScript e C#. Ciò potrebbe riflettere la qualità del codice Java nei dataset di addestramento, che probabilmente contiene un numero maggiore di vulnerabilità rispetto a quello relativo agli altri linguaggi di programmazione. Analizzando l’impatto della dimensione del modello sulle prestazioni, si osserva che il syntax pass rate tende a migliorare all’aumentare delle dimensioni del modello. Tuttavia, il security pass rate rimane pressoché invariato, oscillando intorno a una media del 50%. Conclusione Sebbene gli LLM siano particolarmente abili nella generazione di codice sintatticamente e semanticamente corretto a partire da specifiche in linguaggio naturale, continuano a introdurre vulnerabilità di sicurezza ad un ritmo preoccupante. Una delle cause principali risiede nella scarsa qualità dei dataset di addestramento, che contengono errori e falle di sicurezza. Inoltre, questi modelli faticano ad individuare proprietà strutturali del programma necessarie per correggere o prevenire tali vulnerabilità. Mentre la risoluzione di queste problematiche richiederà tempo e progressi, la generazione di codice tramite Intelligenza Artificiale diverrà sempre più diffusa. Senza un cambio di rotta, le applicazioni che utilizziamo potrebbero contenere sempre più vulnerabilità di sicurezza – spesso senza che nessuno se ne accorga.

The North Korea-linked threat actor known as UNC4899 has been attributed to attacks targeting two different organizations by approaching their employees via LinkedIn and Telegram. "Under the guise of freelance opportunities for software development work, UNC4899 leveraged social engineering techniques to successfully convince the targeted employees to execute malicious Docker containers in their respective workstations," Google's cloud division said [PDF] in its Cloud Threat Horizons Report for H2 2025. UNC4899 overlaps with activity tracked under the monikers Jade Sleet, PUKCHONG, Slow Pisces, and TraderTraitor. Active since at least 2020, the state-sponsored actor is known for its targeting of cryptocurrency and blockchain industries. Notably, the hacking group has been implicated in significant cryptocurrency heists, including that of Axie Infinity in March 2022 ($625 million), DMM Bitcoin in May 2024 ($308 million), and Bybit in February 2025 ($1.4 billion). Another example that highlights its sophistication is the suspected exploitation of JumpCloud's infrastructure to target downstream customers within the cryptocurrency vertical. According to DTEX, TraderTraitor is affiliated with the Third Bureau (or Department) of North Korea's Reconnaissance General Bureau and is the most prolific of any of the Pyongyang hacking groups when it comes to cryptocurrency theft. Attacks mounted by the threat actor have entailed leveraging job-themed lures or uploading malicious npm packages, and then approaching employees at target companies with a lucrative opportunity or asking them to collaborate on a GitHub project that would then lead to the execution of the rogue npm libraries. "TraderTraitor has demonstrated a sustained interest in cloud-centric and cloud-adjacent attack surfaces, often with a final goal of compromising companies that are customers of cloud platforms rather than the platforms themselves," cloud security firm Wiz said in a detailed report on TraderTraitor this week. The attacks observed by Google Cloud targeted the respective organizations' Google Cloud and Amazon Web Services (AWS) environments, paving the way for a downloader called GLASSCANNON that's then used to serve backdoors like PLOTTWIST and MAZEWIRE that can establish connections with an adversary-controlled server. In the incident involving the Google Cloud environment, the threat actors have been found to employ stolen credentials to interact remotely using Google Cloud CLI over an anonymous VPN service, carrying out extensive reconnaissance and credential theft activities. However, they were thwarted in their efforts due to the multi-factor authentication (MFA) configuration applied to the victim's credentials. "UNC4899 eventually determined the victim's account had administrative privileges to the Google Cloud project and disabled the MFA requirements," Google said. "After successfully gaining access to the targeted resources, they immediately re-enabled MFA to evade detection." The intrusion targeting the second victim's AWS environment is said to have followed a similar playbook, only this time the attackers used long-term access keys obtained from an AWS credential file to interact remotely via AWS CLI. Although the threat actors ran into access control roadblocks that prevented them from performing any sensitive actions, Google said it found evidence that likely indicated the theft of the user's session cookies. These cookies were then used to identify relevant CloudFront configurations and S3 buckets. UNC4899 "leveraged the inherent administrative permissions applied to their access to upload and replace existing JavaScript files with those containing malicious code, which were designed to manipulate cryptocurrency functions and trigger a transaction with the cryptocurrency wallet of a target organization," Google said. The attacks, in both cases, ended with the threat actors successfully withdrawing several million worth of cryptocurrency, the company added. The development comes as Sonatype said it flagged and blocked 234 unique malware npm and PyPI packages attributed to North Korea's Lazarus Group between January and July 2025. Some of these libraries are configured to drop a known credential stealer referred to as BeaverTail, which is associated with a long-running campaign dubbed Contagious Interview. "These packages mimic popular developer tools but function as espionage implants, designed to steal secrets, profile hosts, and open persistent backdoors into critical infrastructure," the software supply chain security firm said. "The surge of activity in H1 2025 demonstrates a strategic pivot: Lazarus is now embedding malware directly into open source package registries, namely npm and PyPI, at an alarming rate." Update Veracode, in a new analysis published on July 31, 2025, said it uncovered a dozen malicious packages published by North Korean threat actors as part of a cryptocurrency stealing campaign. The malware is assessed to be a variant of BeaverTail, capable of harvesting system information, exfiltrating data from cryptocurrency wallet apps and browser extensions, and downloading additional Python payloads. None of the identified packages are available for download from npm. (The story was updated after publication on August 2, 2025, to include additional insights from Veracode.)