washington university

Cyber Command turns to veteran of intelligence agencies for top AI role U.S. Cyber Command has a new artificial intelligence leader, as the country’s military races to integrate the transformative technology into digital operations. Ronzelle Green, who most recently served as the head of research and development at the National Geospatial-Intelligence Agency (NGA), has been tapped as the command’s latest chief AI officer. He replaces Brig. Gen. Reid Novotny, who was the first appointee to the role, which Cyber Command created last year to steer its portion of a Pentagon-wide push to adopt AI. For instance, in fiscal 2026 the command requested just $5 million for an “AI for Cyber Operations” effort. The request mushroomed to $138 million in fiscal 2027, according to budget documents. “To maintain decision superiority in this environment, USCYBERCOM must field AI capabilities that allow cyber operators to process large volumes of data, identify malicious activity, and respond to threats faster than human operators alone can achieve,” the documents state. Prior to NGA — a combat support agency within the Defense Department that is also an element of the U.S. intelligence community — Green was the chief information officer for the Defense Counterintelligence and Security Agency. He was also the director of Commonwealth Integration in the Office of the Under Secretary of Defense for Intelligence and Security, where he was responsible for incorporating Five Eyes partners into the broader defense intelligence enterprise. Green is also a senior U.S. Coast Guard Reserve officer. He earned a bachelor of science degree from Oral Roberts University, a master’s degree from North Carolina State University, a master’s in Strategic Intelligence from the National Defense Intelligence College and has a Ph.D in Systems Engineering from The George Washington University. “U.S. Cyber Command welcomes Rear Adm. Ronzelle Green as the Command’s new Chief Artificial Intelligence Officer and thanks Brig. Gen. Reid Novotny for his leadership and service,” a command spokesperson said in a statement. Unlike the National Security Agency — which shares resources with Cyber Command at Fort Meade, Maryland, and established its own Artificial Intelligence Security Center in 2023 – it’s unclear to current and former officials where Cyber Command’s AI focal point is. Pilot programs and other efforts are sprinkled throughout the entire military organization and its forces. “It’s everywhere. Everyone is doing their own ‘home brew’ thing,” according to a command official who was not authorized to speak on the record. Multiple sources, speaking on the condition of anonymity, said that one of Green’s priorities should be to consolidate and organize the various efforts underway. “He needs to be the one-stop shop for AI,” one of these people said. Martin Matishak is the senior cybersecurity reporter for The Record. Prior to joining Recorded Future News in 2021, he spent more than five years at Politico, where he covered digital and national security developments across Capitol Hill, the Pentagon and the U.S. intelligence community. He previously was a reporter at The Hill, National Journal Group and Inside Washington Publishers.

License plate cameras may be next target after Supreme Court reins in location tracking A recent Supreme Court ruling which found that cell phone location history searches require a warrant could have a ripple effect on the use of automated license plate readers (ALPRs), some legal scholars and ALPR opponents say. Last month’s Chatrie v. United States ruling — the first major Supreme Court Fourth Amendment case in eight years — addressed fundamental questions about law enforcement’s use of private data. Over the past several years, police have been pressuring big tech firms to turn over users’ location history data so that they can pinpoint which phones were in the area of a crime scene during a particular time period. The Supreme Court ruling suggests these “geofence” searches are protected by the Fourth Amendment, and opens the door to a debate about potential parameters for how police can use them in the future. If a warrant is ultimately needed for ALPR searches, experts say, it would radically limit how the networks of cameras can be used and would change modern policing. Flock Safety, the country’s leading ALPR vendor, has between 90,000 to 100,000 cameras installed on public roadways and collects data on about 20 billion license plates a month, it says. Police increasingly use the data collected by these cameras to identify suspects. Although the Chatrie case involved Google — which gave police a list of the phones inside the so-called geofence where a crime was committed — there are similarities with ALPR data. Michael Soyfer, an attorney at the Institute for Justice, said at a Wednesday briefing that the Supreme Court emphasized that the “retrospective and indiscriminate” nature of the location history surveillance was particularly worrisome — adjectives which also apply to ALPR data. The Chatrie ruling also bolsters the case against ALPRs because the court focused on what the police had access to globally in their database, according to Soyfer. “The justices drilled down on what was in the database and not just what police happened to access at a point in time,” he said. “The court's really emphasizing that it's looking at the capabilities of the technology overall rather than just what police did with it.” The Chatrie decision could have implications for cases involving reverse keyword searches, cell tower dumps and law enforcement purchase of commercial location data from brokers, Soyfer said. A spokesperson for Flock Safety said in a statement that the court decision is not relevant to the company. “The Supreme Court’s decision addresses geofence warrants for Google location history, which is categorically different from license plate recognition technology,” the statement said. “Google location history involves data from a person’s own mobile device and reveals continuous movements across both public and private places.” “Flock’s ALPR technology, by contrast, captures point-in-time images of vehicles in public view.” Courts have “repeatedly and uniformly” treated ALPR differently from technologies like cell site location information and mobile geolocation data, according to the statement. “We do not believe this ruling undermines that settled precedent,” the statement said. A side note in the Supreme Court opinion appears to differentiate the standard for what qualifies as a Fourth Amendment search based on whether the tracking occurs on “public roads,” according to the statement. ‘Indisputably private’ In the Chatrie case, the government argued that because the geofence searches involved the collection of data from short timeframes they did not warrant Fourth Amendment protections. The high court sided with Chatrie, saying that the government is “wrong about the incapacity of short-term location information to reveal private matters.“ Even only a few hours worth of location history can reveal whether a citizen went to an “indisputably private” place such as the psychiatrist, an abortion clinic, an AIDS treatment center, a strip club, or a by-the-hour motel, according to the opinion. “Location History enables police officers to focus on precisely those sites — to see, in a given time block, who shows up,” it said. The fact that license plates are connected to law enforcement networks of personal data that reveal location, travel patterns and where people live make ALPR searches similar to cell phone location history searches in important ways, said Andrew Guthrie Ferguson, a law professor at George Washington University and the author of Your Data Will Be Used Against You. “Modern ALPRs are just the connecting point to a much larger system of personally revealing information stored in police and connected public databases,” he said in an interview. Those databases often contain not just ALPR data, but also documentation of individuals’ social media activity, surveillance video from thousands of public and private cameras, body cameras and drone video footage, gunshot audio detection sensor data and police dashboard cameras, Ferguson said. Police also store granular information about people they are tracking in the databases, allowing them to build extraordinarily detailed dossiers. “There are some differences with the nature of license plates that are after all designed for identification, but I think it is a mistake to think about ALPRs standing alone,” Ferguson said. “Chatrie certainly strengthens the Fourth Amendment case against the warrantless collection of ALPR data.” Suzanne Smalley is a reporter covering digital privacy, surveillance technologies and cybersecurity policy for The Record. She was previously a cybersecurity reporter at CyberScoop. Earlier in her career Suzanne covered the Boston Police Department for the Boston Globe and two presidential campaign cycles for Newsweek. She lives in Washington with her husband and three children.

Justices rule that cellphone location histories are protected by the Fourth Amendment The Supreme Court on Monday ruled that police use of location history data pulled from individuals’ phones by tech companies should be considered a Fourth Amendment search and requires a warrant. The decision is a significant win for privacy and civil liberties advocates who have long held that so-called geofence searches are unconstitutional or at least should be more heavily scrutinized and narrowed. Geofence searches occur when police ask tech companies to produce a list of every phone located in a particular area during a specific time frame, allowing police to pinpoint potential suspects when they have none. The case centers on Okello Chatrie, a Virginia man who was charged with bank robbery in 2019 after police nabbed him by using Google location history records. While police had a warrant for the search in this case, a lower court ruled that a warrant wasn’t needed at all. Chatrie’s lawyers challenged that contention, arguing that the search was unreasonable and should not be allowed even with a warrant. The high court on Monday sent the case back to the U.S. Court of Appeals for the Fourth Circuit for a ruling on whether the Chatrie warrant was reasonable, sufficiently narrowly tailored and otherwise constitutional. The appeals court could now set strict parameters for when and how geofence warrants are executed. In addition to establishing that geofence searches require warrants, the court also was skeptical of the government’s position that its search was protected by third-party doctrine, an argument law enforcement has made to defend geofence searches, asserting that because individuals choose to share information with Google it is no longer private. “The big deal is that the court is saying that when a person uses their cellphone in the way people in modern society use a phone, they don't forfeit their Fourth Amendment rights,” Greg Nojeim, director of the Security and Surveillance Project at the Center for Democracy and Technology, said in an interview. “The court also dealt a strong blow against the legal doctrine that says if you share information with a third party, such as Google, Apple or Microsoft, the government can compel the disclosure of your information without a warrant,” Nojeim added. “The court is removing brick by brick the foundation of this third-party doctrine.” While Google no longer stores location history, making it impossible for law enforcement to obtain it, a host of other companies track and store user location data, including Uber, Lyft and Apple. The decision therefore will have a big impact on law enforcement, which sometimes does not use warrants to compel big tech companies to disclose such data. Had the court ruled that geofence searches are not protected by the Fourth Amendment, experts predicted a surge in law enforcement use of the tool as well as more aggressive use of other kinds of reverse searches such as keyword inquiries to see lists of names searching specific terms on Google. “Chatrie is a major win for Fourth Amendment privacy,” George Washington University law professor Andrew Guthrie Ferguson said via email. “In requiring a warrant to obtain location data from your cellphone, the Supreme Court upgraded the Fourth Amendment to meet the digital age.” Ferguson, also the author of the book Your Data Will be Used Against You, added that the court’s flat rejection of the government’s argument that location data can’t be collected without a judicial warrant sends a “clear signal that it treats digital privacy seriously.” 'Inquisitive eyes of the government' In a 6-3 vote, the justices held that searching someone’s Google location history is akin to searching private papers and journals, which is expressly prohibited under the Fourth Amendment absent a warrant. “Google users regularly employ Location History as a personal journal,” Justice Elena Kagan wrote for the majority. “In that way, Location History resembles other private materials — e.g., emails, documents, photographs, or calendars — that even if stored on Google’s servers, a user reasonably views as his own and expects to be shielded from the ‘inquisitive eyes’ of the government.” The opinion noted that Google repeatedly prompts users to turn on location history and even cautions Android users that their phone will not “work correctly” with it turned off. Location history also remains active regardless of whether the user has a Google app open and even when a phone is not in use, the opinion said, noting that even if the user deletes the app for which they turned on location history, Google continues to capture movement history unless a user “affirmatively stops it.” The precision of location history searches as well as the fact that police can make them retroactively is also an important consideration, the opinion said, noting that location history pinpoints location within around twenty meters, records it about every two minutes and can even see what floor of a building an individual is on by capturing elevation. “The exposure of that information to Google is merely what happens when a user avails himself of one of the services on his cell phone,” the opinion said. “The Government’s argument that generating Location History… is a voluntary choice is meritless. That argument ignores how and why Google users turn on Location History.” General warrants Chatrie’s lawyer had argued for the court to reach a more expansive decision, outlawing geofence searches entirely. Geofence searches should be considered general warrants, which are unconstitutional, Chatrie’s lawyer, Adam Unikowsky, told Recorded Future News in May. A general warrant allows the police to search large groups of people for evidence of a crime without probable cause. Before the Declaration of Independence was written, the authorities used general warrants to search private papers of large numbers of people, usually searching for records relating to tax payments. General warrants don’t name places to be searched or things to be seized. They simply allow police the authority to “rummage through people's possessions,” Unikowsky said. Geofence searches that allow police to search large numbers of phone owners’ location histories, exposing people who have nothing to do with a crime to police scrutiny, constitutes a general warrant, Unikowsky said. “We understand the warrant to search every single person’s Google account, and we're talking about tens of millions here, is akin to a general warrant, which authorized the search of thousands of houses for evidence,” Unikowsky said. On Monday, Unikowsky told Recorded Future News the ruling “persuasively explains why obtaining Location History is a Fourth Amendment search. The Court’s decision marks a major victory for digital privacy.” Suzanne Smalley is a reporter covering digital privacy, surveillance technologies and cybersecurity policy for The Record. She was previously a cybersecurity reporter at CyberScoop. Earlier in her career Suzanne covered the Boston Police Department for the Boston Globe and two presidential campaign cycles for Newsweek. She lives in Washington with her husband and three children.

Supreme Court signals location data searches should require a warrant The Supreme Court signaled during oral arguments Monday that it is likely to rule that police sweeps of all cell phones located in an area surrounding a crime scene qualify as a Fourth Amendment protected search and therefore require a warrant. While police now typically get warrants to execute so-called geofence searches, the government is arguing in the Chatrie v. United States case that such probes do not trigger Fourth Amendment protections. Privacy advocates had worried that the high court would rule that geofencing does not qualify as a constitutionally protected search, opening the door to much broader use of warrantless reverse searches of all types, including those where police ask Google to turn over information about everyone who searched for a specific keyword. By the end of Monday’s hearing it seemed clear that a majority of justices will rule that location data searches require a warrant. Some justices seemed to indicate that they see a need to rule in a way that lays out requirements for location data searches to be as narrow as possible. The justices did not split across typical ideological lines. Justice Neil Gorsuch, a conservative, joined liberal justice Sonia Sotomayor in asking the government’s lawyer hard questions about whether he was arguing that police should also be able to seize other data like photos, emails and Google documents without a warrant. The case is the first Supreme Court argument to deal with data privacy since 2018, when the court ruled that police need a warrant to obtain 7 days or more of an individual's historic cell-site location information — gathered from cell towers — in the landmark Carpenter v. United States case. Okello Chatrie, the plaintiff in the current case, is serving a nearly 12-year-sentence for bank robbery. He was only convicted because police were able to find him after they asked Google to turn over the names of everyone whose cell phone location data showed them in the area surrounding the credit union that was robbed in the time frame during which the robbery occurred. Chatrie’s lawyer, Adam Unikowsky, argued that geofence searches should be considered general warrants, which are banned under the Fourth Amendment. General warrants gave British officers in the colonial era blanket authority to search without probable cause. Some justices on Monday seemed skeptical of that argument, but still raised concerns about how broad geofence searches can be. Google submitted an amicus brief on behalf of Chatrie, arguing that geofence searches are often broadly worded and invasive. The tech giant told the court that it has received, for example, a geofence warrant that covered several search areas for a combined 2.5 square miles of San Francisco for more than 48 hours worth of time. “This geofence warrant would have exposed the location history of thousands of users,” the Google brief said. In another case, Google said, police asked for data that would have covered more than 3,000 users located in a 489-acre area that included the Islamic Center of New Mexico, 50 homes, 30 apartment buildings, a park, major roadways and a chunk of the University of New Mexico campus. While the court does not seem poised to ban geofence warrants, the justices did seem to focus attention on how “particularized a warrant must be,” said Andrew Guthrie Ferguson, a law professor at George Washington University and the author of Your Data Will Be Used Against You. “As we have seen in the Supreme Court’s other digital surveillance cases, the Justices seem to be more comfortable with a compromise approach, requiring judicial warrants to put some limits on law enforcement searches, but also not banning them completely,” Ferguson said via email. Ferguson said he sees the court’s apparent inclination to require warrants for geofence searches as a big win since there is currently no mandate for them and police have typically sought warrants as a “concession that they think a court might require one.” “That is why this case is a big deal,” Ferguson said. “If the government's argument wins, police can get a geofence warrant without any probable cause, for any reason, against anyone, anywhere in America.” Suzanne Smalley is a reporter covering digital privacy, surveillance technologies and cybersecurity policy for The Record. She was previously a cybersecurity reporter at CyberScoop. Earlier in her career Suzanne covered the Boston Police Department for the Boston Globe and two presidential campaign cycles for Newsweek. She lives in Washington with her husband and three children.

A congressional subcommittee on Thursday held a roundtable discussion on the potential of artificial intelligence, which took a turn toward the existential as each lawmaker aired their anxieties about the rapidly evolving technology. Rep. James Walkinshaw, D-Va., expressed alarm that federal workers may be using AI chatbots to handle sensitive government data. Rep. William Timmons, R-S.C., asked whether it should be illegal for AI systems to use someone’s likeness to create pornographic images. Rep. John McGuire, R-Va., expressed concerns that AI systems could deny U.S. military forces from taking lethal actions due to a model’s conclusion for “moral” behavior. And Rep. Yassamin Ansari, D-Ariz., raised concerns about the Trump administration’s use of AI in the war with Iran, the technology’s intensive energy usage and its potential effects on the climate. While members of Congress elsewhere debated other major topics — the scope of the federal government’s surveillance powers, the war with Iran and funding the Department of Homeland Security — the House Oversight Committee’s subcommittee roundtable on “Artificial Intelligence and American Power” brought executives of AI firms, academics and those implementing AI at major companies together with lawmakers. Thursday’s discussion comes as leaders on Capitol Hill grapple with the dizzying pace of global developments in which technology plays a central role. But the conversation quickly considered the potential for artificial intelligence to dwarf every other challenge facing the country. “People in our districts across this country are going to start feeling impacts very soon, and if we don’t start thinking properly and aggressively and proactively about the challenges that AI creates, I fear that we’re going to have a revolution on our hands,” said Rep. Dave Min, D-Calif. The subcommittee’s ranking Democrat, Rep. Maxwell Frost of Florida, also expressed optimism about AI’s ability to cure diseases and boost the economy. But Frost, currently the youngest member of Congress, worried that the technology would outpace lawmakers and pose potentially disastrous consequences if not addressed early on. “I don’t have faith in this institution to actually put the common sense guardrails in place. And then we fast forward ten years, and the house is on fire,” said Frost. “That won’t be good for anybody, whether it’s the industry or working families and people, or this institution itself.” Rep. Eric Burlison, R-Mo., kicked off the meeting with praise for the industry and marveled at how one panelist’s company used AI to automate and fast-track manufacturing in the firm’s factories. “It’s truly like the closest thing to Star Trek I’ve ever seen,” Burlison said. He later inquired about what congressional districts should do to attract AI firms for business. Many also openly fretted about disclosures from technology firms like Anthropic, which recently announced that its Mythos AI model, which the company claims has capabilities so powerful that it is limiting its use to select customers because of its apparent ability to bypass traditional cybersecurity and hack major institutions like banks, government agencies and major corporations. “I recognize AI is not going anywhere,” said Rep. Eli Crane, R-Ariz., a former Navy SEAL who served in combat. “That being said, does anyone on this panel feel or believe, in any way, that as we are going down the road in this AI race, we might be simultaneously engineering our own destruction?” The assembled experts and industry leaders all highlighted AI’s vast and growing capabilities. They urged lawmakers, alongside their policy recommendations, to be thoughtful and well-informed when making policy. Mark Beall, president of government affairs at the AI Policy Network Inc. and a former Pentagon official, warned that Congress risked the country losing its competitive edge on AI if it did not act on key national security concerns. “I don’t think it’s going to kill us,” Robert Atkinson, founder of the Information Technology and Innovation Foundation, a technology think tank, told lawmakers. “At the same time, I do think it’s important for the federal government to seriously fund AI safety research,” Atkinson continued. “We need to know a lot more about how the models work.” Spencer Overton, a George Washington University law professor, said the incentives for AI companies “are really what they should be” when asked by lawmakers whether the firms were good actors. “Constituents are looking for you, not for companies, to step up and protect them,” Overton said. “They’re trusting you, the person that they voted for, to do that, as opposed to companies. That’s the way the system works, right?” Related: OpenAI Widens Access to Cybersecurity Model After Anthropic’s Mythos Reveal

Pa. high court rules that police can access Google searches without a warrant The Pennsylvania Supreme Court ruled Tuesday that police did not need a warrant to obtain a convicted rapist’s Google searches when investigating the crime. In its opinion, the court said that internet users making searches have no reasonable right to privacy because “it is common knowledge that websites, internet-based applications, and internet service providers collect, and then sell, user data.” The case only creates legal precedent in Pennsylvania, but an expert predicted that the ruling will lead more police departments to feel confident about warrantless searches for internet queries. “If a rather progressive state like Pennsylvania gives the green light to warrantless collection of your search queries, I think it fair to say that is going to open up its use across the nation,” said Andrew Ferguson, a law professor at George Washington University and the author of the forthcoming book “Your Data Will Be Used Against You.” Police in the case before the court had hit a dead end in their probe of a rape. As a last attempt to find the rapist, they asked Google to produce a list showing anyone who had searched for the victim’s address in the week before the rape and home invasion occurred. Google found a hit and told police that someone at an IEP address tied to the home of the defendant in the case, John Edward Kurtz, had looked up the victim’s address a few hours before the crime took place. The court noted that Google’s privacy policy is explicit about the fact that it will share search histories with third parties. “In the case before us, Google went beyond subtle indicators,” the opinion says. “Google expressly informed its users that one should not expect any privacy when using its services.” The court also said that internet users have the ability to not expose their data by using different methods for finding information. “The data trail created by using the internet is not involuntary in the same way that the trail created by carrying a cell phone is,” according to the opinion. Ferguson said the decision is an important development on a hotly contested and developing legal question. Allowing police to access Google searches without a warrant creates a “chilling environment,” he said, pointing out that many people ask Google questions they wouldn’t ask even their spouses. “The danger of a reverse keyword search is that it allows the police to rummage through our digital questions and queries and thus, by inference, our minds,” Ferguson said. “A list of your questions to Google is a direct link to your thinking, and we usually try to prevent the government from having access to that information.” Suzanne Smalley is a reporter covering digital privacy, surveillance technologies and cybersecurity policy for The Record. She was previously a cybersecurity reporter at CyberScoop. Earlier in her career Suzanne covered the Boston Police Department for the Boston Globe and two presidential campaign cycles for Newsweek. She lives in Washington with her husband and three children.

Law enforcement is using AI to synthesize evidence. Is the justice system ready for it? Max Dorsey, a small-town South Carolina sheriff probing a sprawling conspiracy case, has been using a little known artificial intelligence tool to augment his investigative team. In an effort to process the vast amount of data his agency is collecting and save investigators time, Dorsey has turned to TimePilot, software produced by the startup Tranquility AI. The platform is now being used by at least a dozen law enforcement agencies nationwide. “The tool allows us to sort through massive amounts of data that the human brain just cannot process because it's so much,” Dorsey said. “It’s not unusual to find a cell phone that has a terabyte of data and it is very difficult for a person to properly look through all that.” The sheriff, whose department is located in largely rural Chester County, says he counts on TimePilot to summarize key elements of the conspiracy case and quickly surface relevant evidence. Dorsey can type in a phrase, and the AI spits back information pulled from the body of data his team has put into the system. Dorsey said that while he can’t guarantee TimePilot is accurate 100% of the time, he does not rely on the AI’s information without going back to the source evidence and verifying it. Maryland-based Tranquility, which came out of stealth mode in February, is one of a few new companies offering police and prosecutors artificial intelligence that synthesizes evidence to deliver neat summaries and tidily packaged insights. Other companies in the category include Truleo and Allometric. All three firms promise law enforcement they can save valuable time. Unlike Tranquility, Allometric offers little information on its website about how its product works and which agencies use it. Truleo discloses more, saying that investigators can access its “virtual crime analyst” for automated reports, “case query and hardening” and witness canvassing. Digital freedoms groups and advocates for defendants say such technology corrupts the criminal justice system. They point to the potential for exculpatory evidence to be missed when police and prosecutors are relying on AI. The concern is that cops and district attorneys are overworked and may be tempted to rely on AI summaries without combing through case data to see what TimePilot — or similar products like Allometric’s AirJustice and Truleo’s Analyst — have omitted or checking the software’s output against original evidence for accuracy. The advocacy groups also cite the risks of AI hallucinations and sycophancy — instances when the technology tells a user what it thinks they want to hear. Experts and advocates interviewed for this story said they had never heard of artificial intelligence being used to synthesize evidence in criminal cases. Those experts offered assessments after reviewing the numerous claims, case studies and videotaped demos featured on Tranquility AI’s website. “Summarizing pages and pages of evidence or hours of footage … is really just editorializing and, when liberty is at stake, these shortcuts are really dangerous,” said Tom Bowman, a former public defender who is now policy counsel for the Security and Surveillance Project at the Center for Democracy and Technology. “You're creating risks that the AI is going to omit context, mislabel events, even overlook exculpatory evidence, and when that gets incorporated into a narrative of a case that's not just a technical flaw — it’s a civil rights violation.” ‘Human oversight must remain at the center’ Many of the law enforcement agencies deploying TimePilot have only had the tool on hand for a few months. How the AI will be treated by courts remains to be seen — if judges and juries are even apprised that it has been used, which experts said is far from a foregone conclusion. A spokesperson for Tranquility did not respond to multiple requests for comment. The company emphasizes its “commitment to ethical AI development” on its website, saying that TimePilot is designed to “uncover objective facts while actively preventing bias.” The company addresses ethics in a blog post added to the site on September 25, saying that AI “should support, not replace, judicial or law enforcement decision-making. Human oversight must remain at the center of the process.” The blog post acknowledges concerns about expanded access to sensitive personal data through AI use, saying that records must be anonymized when possible and only used for their “intended purpose.” Defendants and their lawyers “must have the ability to question and understand AI-influenced evidence or recommendations,” the blog post says, but does not explain how they can do so. The founder of Allometric, which launched this summer, said the company’s AI is now being piloted by seven jurisdictions, including one public defender’s office, and that despite its newness, it already has been used to secure at least one conviction. “Attorneys remain ultimately responsible for identifying and disclosing exculpatory evidence and complying with their discovery obligations,” Allometric CEO Patrick Robinson said via email. “Our goal is to make the discovery process faster, more accurate, and more consistent, never to automate away professional judgment.” Truleo’s AI product for investigators launched in June. It costs $200 per month for each user. Tejas Shastry, the company’s co-founder and chief technology officer, also placed the onus on investigators to ensure they are backstopping the AI. “The product summarizes all evidence the investigator provides it,” Shastry said via email. “It is up to the investigator how to use those summaries to further the case.” Hundreds of police departments have signed on to use the Truleo product since the June launch, he said. The use of such products will soon become more widespread, according to Ian Adams, a former police officer and criminal justice professor at the University of South Carolina. “This is a new category of AI products that I see a lot of development, commercialization, and promise in,” Adams said via email. Adams also sees major weaknesses, particularly in terms of what he calls the “quiet structural problem of omission,” though he said that hallucinations are unlikely. “The ‘savings’ the vendor promises never fully materialize, because you can’t safely shortcut the due diligence,” said Adams, who has led independent research into the use of AI by law enforcement, including by evaluating Truleo. “While I’m optimistic about this product category, we need a lot more evidence,” he said. The appeal of a time-saver Most of the 12 law enforcement agencies named as clients on Tranquility’s website are in rural areas. The only large entity is the Orleans Parish district attorney's office in New Orleans. In July, the company inked a deal with top government information technology vendor Carahsoft — a relationship that will likely dramatically increase TimePilot’s reach. In the meantime, Tranquility has given some agencies free access through grant awards, according to law enforcement officials interviewed for this story. Sheriff Max Dorsey is one of several U.S. law enforcement officials with early access to an AI platform that helps process evidence. Image courtesy of Max Dorsey It is unclear how much the product costs users who have not been awarded a grant. Dorsey said he negotiated a special price by agreeing to only use TimePilot for one case but declined to say how much he paid. A use case featured in a video on the Tranquility website shows how the AI breaks down evidence from the Boston Marathon bombing investigation for a hypothetical prosecutor. An invisible user types into a search bar, asking, “What investigative leads regarding the travel to Russia need to be followed up on?” TimePilot instantly generates two lists. One offers several examples of “key investigative steps that should have been taken” when investigators probed a key suspect’s travel. A second gives several reasons why TimePilot thinks the travel should be considered “particularly significant.” The AI is not just enumerating facts but is also offering analysis by picking which facts to highlight. The software understands at least 120 languages and most slang, according to the Tranquility website. It also can read and summarize handwritten reports. Investigators can dump data from a variety of sources into the platform, including from Axon, which makes automated license plate readers and police body cameras; Ring, which manufactures doorbell cameras; Cellebrite, which extracts data from cellphones; Cash App; Venmo; Prison Calls; TikTok; Instagram and Facebook, the website says. The platform also can digest data from so-called tower dumps, which provide police with records of every device connected to a specific cell tower during a particular time period, law enforcement officials said. McCord Larsen, a prosecutor in rural Cassia County, Idaho, is using the product for free thanks to a grant. Larsen said he asked TimePilot to provide information on a specific evidentiary question and it zeroed in on relevant information in seconds. “The material I am searching through is thousands of pictures, hours of video and, of course, thousands of pages,” Larsen said via email. TimePilot’s analysis of the question Larsen asked “saves me hours of time,” he said. “I can see where it'd be very useful with cold cases — somebody just coming in and needing, like, a crash course on what went on.” — Kelly Marshall, police chief for Choctaw, Oklahoma Police in Choctaw, Oklahoma, also have recently received a grant to use TimePilot, according to Kelly Marshall, the department’s chief. Tranquility’s demo of the product was impressive, she said. “It can link a lot of clues together,” Marshall said. “I can see where it'd be very useful with cold cases — somebody just coming in and needing, like, a crash course on what went on.” TimePilot’s ability to produce a “quick snapshot” is valuable, she said, because it “cuts to the chase.” Andrew Guthrie Ferguson, a law professor at George Washington University and the author of the forthcoming book “Your Data Will Be Used Against You: Policing in the Age of Self-Surveillance,” said it's no coincidence that TimePilot and similar products are hitting the market now. “Prosecutors will soon be deluged with data from body cams, surveillance cams, and other data-rich surveillance technologies,” he said. “The temptation to upload an overwhelming amount of data into a bespoke AI model will be too strong for many offices to resist.” ‘Bring the receipts and get the pleas’ Tranquility’s website includes videotaped case studies showing TimePilot working with fact sets from well-known cases — including the Jeffrey Epstein probe and the Gabby Petito murder investigation — alongside an article about the despair victims’ loved ones feel when crimes go unsolved. Visitors are repeatedly invited to “book a demo.” One section of the website imagines how TimePilot could have uncovered vital evidence for investigators who spent 13 years desperately trying to find a serial killer terrorizing sex workers on Long Island. The suspect in that case, Rex Huermann, was arrested in 2023 and has been charged with killing seven women. “Make your investigators super human with TimePilot,” the website says. “Process months of data in minutes. … Force multiply your investigative team.” TimePilot can do the work of 10 investigators making an average of $60,000 a year, clearing 50 backlogged cases annually, the website promises. Time saved? 14,520 hours. Cost savings are estimated to be $418,846. In the section of the website aimed at prosecutors, similar promises are made. “Prosecutors are overwhelmed,” the website says, asserting that plea negotiations are cut from 30 days to three on average when TimePilot is used. “Bring the receipts and get the pleas with TimePilot.” Bowman, the former public defender, said that the marketing underscores the dangers. Overworked people will turn to Tranquility and products like it, he said, and those same people may be too busy to check their output. “We might want to think that prosecutors are always going to be able to exercise their discretion and say, ‘Oh, this doesn't seem like it actually matches the initial investigation or other evidence that we have,’” Bowman said, “but the reality is that when you are in a courtroom, both prosecutor and defense attorneys might not have had a good opportunity to fully review the case.” Lawyers on both sides of criminal cases are stretched, to be sure. Another product, JusticeText, was created by technologists interested in helping public defenders keep up with better-resourced prosecutors. The result is an audiovisual evidence management software that creates automated transcripts of body camera footage and interrogation videos and allows for AI-powered evidence review. ‘AI is not trained to be a prosecutor’ "Brady rights," a term drawn from the 1963 Supreme Court decision Brady v. Maryland, require prosecutors to share any evidence that may prove a defendant is innocent or raise questions about the truthfulness of a witness, for example. Brady violations can lead to conviction reversals. But with technology like TimePilot, instances of exculpatory evidence being withheld could increase as such errors become much harder to detect, according to Jumana Musa, director of the Fourth Amendment Center at the National Association of Criminal Defense Lawyers. “AI is not trained to be a prosecutor; it is trained to look for particular things and put them together,” Musa said. “If your idea is this person has done this thing and there's a gun and a red sweatshirt and a blue car, you say ‘Find me all of these elements,’” Musa said. “Maybe what you're missing is something else that is not a gun, a red sweatshirt or a blue car.” It’s not clear how many convictions, if any, have resulted from TimePilot’s use so far. “What is the failsafe? Where's the process? Is there a process?” — Jumana Musa of the National Association of Criminal Defense Lawyers The fact that Tranquility is not publicly revealing how its AI is trained troubles Musa. The ethics blog post on Tranquility’s website acknowledges the issue but doesn’t disclose the training, saying only that AI companies “must be able to explain how AI models work, what data they use, and what limits they carry. This makes accountability possible.” While it is not unusual for a private AI company to guard its proprietary information, Musa said that such secrecy is not appropriate in the criminal justice context. (Patrick Robinson, of Allometric, said the company does not train its own large language models and instead relies on application programming interfaces and models developed by “leading technology companies.”) Ferguson echoed Musa’s point, saying that AI “shifts primary responsibility from a democratically-authorized and licensed lawyer to an undemocratic and unlicensed algorithm.” The public’s perception matters, too, the experts said. Citizens know how to assess fingerprint evidence because the process for collecting them is common knowledge. Independent experts also can replicate fingerprinting based on established science. The same is not true for a large language model using methods protected as trade secrets, Musa said. “What is the failsafe?” she said. “Where's the process? Is there a process?” Tranquility’s website does not address model training in detail. Neither do Truleo’s and Allometric’s. In Tranquility’s case, the website does hint at why its leadership believes its software is vital, however. The preamble to the U.S. Constitution is featured prominently, highlighting the founding document’s promise of “domestic Tranquility” — with Tranquility in bold text. For Musa, a different bedrock of the Constitution comes to mind. The Fifth Amendment promises Americans that no citizen shall be "deprived of life, liberty or property without due process of law." “Somebody's life and liberty is at stake, and that is where the most heightened protections should come in,” Musa said. “That is not where we should be outsourcing the development of the case to somebody's black box AI tool.” Suzanne Smalley is a reporter covering digital privacy, surveillance technologies and cybersecurity policy for The Record. She was previously a cybersecurity reporter at CyberScoop. Earlier in her career Suzanne covered the Boston Police Department for the Boston Globe and two presidential campaign cycles for Newsweek. She lives in Washington with her husband and three children.