The Trump administration has closed a deal with Anthropic to give U.S. intelligence agencies access to the company's artificial intelligence tools. The agreement, finalized in recent days, opens the door for spy agencies to use Anthropic’s models across their operations. Specific terms of the contract have not been made public.
What the deal covers
Under the agreement, Anthropic will provide its AI systems to multiple agencies within the U.S. intelligence community. The exact scope remains unclear — which agencies are involved, how broad the access is, and whether the tools will be used for analysis, surveillance, or other tasks haven't been disclosed. Intelligence officials declined to comment on the record about the arrangement.
The deal marks one of the first formal contracts between a major AI lab and the government’s spy apparatus. Anthropic, a San Francisco-based startup, has positioned itself as a safety-focused alternative to rivals like OpenAI. The company’s models are designed with built-in guardrails meant to prevent harmful uses.
Anthropic’s role in national security
Anthropic has long said it wants its AI to be used responsibly. But critics have raised concerns about handing powerful AI tools to intelligence agencies, where secrecy is standard and oversight can be limited. The company has not published any detailed ethical framework for how its technology might be used in classified settings.
The deal comes as the Trump administration has made AI a priority for national security. Executive orders and policy memos over the past year have pushed agencies to adopt AI faster, with less public debate about risks. Privacy advocates worry that the lack of transparency around this contract could set a precedent for even deeper AI integration into intelligence work.
Broader push for AI in intelligence
Other AI companies have also pursued government contracts. OpenAI, Google, and Meta have all pitched their tools to defense and intelligence clients. But Anthropic’s deal appears to be the first that specifically targets spy agencies rather than military or law enforcement branches.
The intelligence community has been experimenting with AI for years, from sifting through surveillance data to generating reports. But access to cutting-edge large language models like Anthropic’s could change how analysts work. Agency officials have described the technology as a force multiplier, though they've also acknowledged risks around bias, hallucination, and security.
What remains unknown
Key questions are unanswered. Will Anthropic have any ability to monitor how its tools are used inside classified networks? Who inside the company approved the deal, and what concessions did the government demand? The intelligence agencies’ inspector general has not announced any review of the contract.
A spokesperson for the Director of National Intelligence said the office does not comment on specific contracts. Anthropic declined to provide details beyond a brief statement confirming the agreement was complete.
The deal is done. But the full list of terms and the agencies involved likely won't be known until Congress or an oversight body pushes for answers.
