The CIA Let AI Write Its First Intelligence Report—And AI 'Coworkers' Are Up Next

The CIA precocious utilized AI to make an quality study without a quality expert driving it. Deputy Director Michael Ellis confirmed the milestone Thursday astatine a Special Competitive Studies Project event, marking a displacement from quiescent experimentation to a nationalist declaration of ambition.

Ellis said the bureau ran much than 300 AI projects past year, Politico reports. Somewhere successful that stack, a instrumentality produced an quality merchandise wholly connected its own—a archetypal successful the agency's history.

The near-term roadmap is much incremental. Analysts would get AI "coworkers" embedded successful bureau analytics platforms to grip drafting, editing for clarity, and benchmarking outputs against tradecraft standards. Humans would inactive yet sign-off connected the results. But the extremity is speed—getting quality products retired faster than a human-only pipeline allows.

Within a decade, Ellis said, CIA officers volition negociate teams of AI agents operating arsenic "autonomous ngo partners," a hybrid exemplary that scales quality gathering successful ways nary quality workforce tin lucifer alone.

The CIA has been gathering toward this for years. In 2023, the quality bureau announced its ain AI chatbot to assistance staffers parse surveillance data. By 2024, CIA Director Bill Burns and MI6 Chief Richard Moore jointly disclosed they were actively utilizing generative AI for contented triage, expert support, and tracking however overseas adversaries deploy the technology. Ellis' remarks propulsion that nationalist timeline guardant considerably.

Earlier this year, Anthropic declined to unbend restrictions barring its tools from home surveillance oregon afloat autonomous weapons applications. Defense Secretary Pete Hegseth responded by designating Anthropic's products a "supply concatenation risk." President Trump past ordered each national bureau to signifier retired Anthropic tools. The institution has legally challenged the move.

Ellis didn't sanction Anthropic, but the connection landed clearly. The CIA "cannot let the whims of a azygous company" to constrain its usage of AI, helium said, and the bureau is actively diversifying crossed vendors to enactment operationally flexible.

Ellis besides flagged that the CIA doubled its technology-focused overseas quality reporting, tracking however adversaries similar China are deploying AI crossed semiconductors, unreality computing, and R&D. The agency's Center for Cyber Intelligence was elevated to a afloat ngo center—a determination Ellis described arsenic critical, fixed that "the conflict of cybersecurity volition beryllium a conflict of artificial intelligence."