Governments adopt AI assistants

The Central Intelligence Agency has started deploying artificial intelligence tools for analysts, while police departments are putting similar software into everyday investigations. (politico.com) (axios.com) Central Intelligence Agency Deputy Director Michael Ellis said on April 9 that the agency recently used artificial intelligence to generate its first intelligence report and plans to build artificial intelligence “co-workers” into all analytic platforms within the next few years. He said the tools will help draft judgments, test conclusions and spot trends, with humans keeping final authority. (politico.com) (nextgov.com) Ellis said the Central Intelligence Agency ran more than 300 artificial intelligence projects in 2025 and expects officers, within a decade, to manage teams of software agents as “autonomous mission partners.” The agency described the shift as a way to speed up analysis built from spy reporting and technical collection. (nextgov.com) (politico.com) Police agencies are using the same basic idea for a different problem: too much evidence for too few investigators. Axios reported in January that departments are using artificial intelligence to sift large stores of video, documents and case files in cold cases, missing-person investigations and trial preparation. (axios.com) In California, GovTech reported on April 1 that San Francisco startup Longeye is negotiating 20 contracts nationwide for software that ingests cell phone data, jail calls, emails, images, spreadsheets and Global Positioning System records into a searchable case file. In one San Mateo County homicide case, the company said its system reviewed 537 jail calls and flagged one key exchange. (govtech.com) In Ohio, Spectrum News said the Euclid Police Department is using an “artificial intelligence crime center” that lets officers search city and partner cameras for objects in real time from patrol cars and phones. Captain David Olszewski said the system does not use facial recognition and requires a reason for each search. (spectrumnews1.com) State lawmakers and federal officials have been trying to draw lines around those tools as adoption spreads. The National Conference of State Legislatures said law enforcement agencies are using artificial intelligence for identification and surveillance, forensic analysis, predictive policing and risk assessment, while governance rules are still “in its infancy.” (ncsl.org) (counciloncj.org) The Department of Justice’s December 3, 2024 report warned that artificial intelligence in criminal justice raises technical, operational and civil-rights concerns across identification, surveillance, predictive policing and forensic analysis. The report said agencies need validation, oversight, transparency and human review before relying on those systems in decisions that affect liberty. (justice.gov 1) (justice.gov 2) Those warnings are not abstract. In June 2024, Detroit agreed to a legal settlement after the wrongful arrest of Robert Williams, a Black man misidentified by facial recognition in a 2020 shoplifting case, and the agreement required new training, audits and limits on how police can use the technology. (michiganpublic.org) (law.umich.edu) Across both intelligence and policing, officials are making the same promise as they wire artificial intelligence into core government work: the software can sort, draft and flag, but a human is supposed to make the call. (politico.com) (justice.gov)