Dario Amodei talked about Thursday that Anthropic plans to notify the Division of Protection’s possibility to mark the AI firm a present-chain possibility in court, a designation he has referred to as “legally unsound.”
The observation comes a number of hours after the DOD formally designated Anthropic a present-chain possibility following a weeks-long dispute over how out of the ordinary retain watch over the army will ought to possess over AI programs. A present-chain possibility designation can bar a firm from working with the Pentagon and its contractors. Amodei drew a firm line that Anthropic’s AI is perchance no longer former for mass surveillance of American citizens or for totally self sustaining weapons, nonetheless the Pentagon believed it will most likely per chance well ought to possess unrestricted discover entry to for “all correct applications.”
In his observation, Amodei talked about the overwhelming majority of Anthropic’s prospects are unaffected by the offer-chain possibility designation.
“With admire to our prospects, it it appears to be like that applies most interesting to the usage of Claude by prospects as a dispute segment of contracts with the Division of War, no longer all reveal of Claude by prospects who possess such contracts,” he talked about.
As a preview of what Anthropic will seemingly argue in court, Amodei talked about the Division’s letter labeling the firm a present-chain possibility is slim in scope.
“It exists to protect the governmentrather than to punish a provider; truly, the law requires the Secretary of War to make reveal of the least restrictive way wanted to enact the goal of retaining the offer chain,” Amodei talked about. “Even for Division of War contractors, the offer chain possibility designation doesn’t (and can’t) restrict uses of Claude or industry relationships with Anthropic if these are unrelated to their insist Division of War contracts.”
Amodei reiterated that Anthropic had been having productive conversations with the DOD over the closing several days, conversations that some suspect got derailed when an inner memo he despatched to workers develop into leaked. In it, Amodei characterized rival OpenAI’s dealings with the Division of Protection as “safety theater.”
Techcrunch event
San Francisco, CA
|
October 13-15, 2026
OpenAI has signed a deal to work with the DOD in Anthropic’s region, a pass that has sparked backlash amongst OpenAI workers.
Amodei apologized for the leak in his Thursday observation, claiming that the firm did not intentionally share the memo or dispute anybody else to contend with out so. “It is some distance never any longer in our interest to escalate the notify,” he talked about.
Amodei talked about the memo develop into written inner “a number of hours” of a assortment of announcements, including a presidential Fact Social put up announcing Anthropic would possibly perchance well be eliminated from federal programs, then Protection Secretary Pete Hegseth’s present-chain possibility designation, and at closing the Pentagon’s deal announcement with OpenAI. He apologized for the tone, calling it “a sophisticated day for the firm” and talked about the memo didn’t replicate his “cautious or knowing of views.” Written six days within the past, he added, it’s now an “out-of-date evaluate.”
He accomplished by announcing Anthropic’s top precedence is to be determined American soldiers and nationwide security consultants retain discover entry to to major instruments within the heart of ongoing most major strive in opposition to operations. Anthropic is currently supporting a number of of the U.S.’s operations in Iran, and Amodei talked about the firm would continue to give its models to the DOD at “nominal mark” for “so long as wanted to homicide that transition.”
Anthropic would possibly perchance well notify the designation in federal court, seemingly in Washington, nonetheless the law insensible the likelihood makes it harder to contest on epic of it limits the identical old programs companies can notify govt procurement choices and provides the Pentagon mountainous discretion on nationwide security matters.
Or as Dean Ball — a vulnerable Trump-period White House adviser on AI who has spoken out in opposition to Hegseth’s remedy of Anthropic — set up it: “Courts are barely reluctant to 2nd-guess the governmenton what’s and is never always a nationwide security topic … There’s a truly excessive bar that one needs to certain in suppose to contend with out that. On the opposite hand it’s no longer impossible.”
Rebecca Bellan is a senior reporter at TechCrunch where she covers the industry, policy, and rising trends shaping synthetic intelligence. Her work has also looked in Forbes, Bloomberg, The Atlantic, The Day-to-day Beast, and somewhat a great deal of publications.
You perchance can contact or review outreach from Rebecca by emailing rebecca.bellan@techcrunch.com or by technique of encrypted message at rebeccabellan.491 on Signal.
