An OpenAI employee who joined dozens of industry peers in filing an amicus brief supporting Anthropic in its legal battle against the Pentagon said he is withdrawing his signature.
In a post on X, Brian Fioca cited Anthropic’s decision to reverse a policy that had limited competitors’ use of its new AI model.
“In light of Anthropic’s policy decision, I am withdrawing my amicus brief signature. I can’t truthfully argue they’re not a supply chain risk,” Fioca, a member of OpenAI’s Go to Market Staff, wrote.
In March, more than 30 employees from Google, OpenAI and Google DeepMind filed an amicus brief supporting Anthropic’s lawsuit against the U.S. Department of Defense (DoD).
Amicus briefs — “friend of the court” filings — allow non-parties with a strong interest in a case’s outcome to provide additional expertise or arguments for judicial consideration.
The brief urged the court to grant Anthropic’s requested relief and concludes: “Until a legal framework exists to contain the risk of deploying frontier AI systems, the ethical commitments of AI developers and their willingness to defend those commitments publicly are not obstacles to good governance or innovation. They are contributions to it. The Court should say so.”
Anthropic Move Rewrites Its Own Support Coalition
Earlier this week, Anthropic reversed course on a controversial safeguard that would have quietly hindered competitors from using its newest AI model, Claude Fable 5, to advance their own systems, Wired reported. Anthropic ultimately abandoned the approach after researchers and industry observers criticized the lack of transparency around the restrictions.
“We’re changing Fable 5’s safeguards for frontier LLM development to make them visible,” Anthropic told Wired in a statement. “We made the wrong tradeoff and we apologize for not getting the balance right.”
Claude Fable 5 included a set of enhanced protections intended to curb high-risk uses of advanced AI. Some of those measures aligned with Anthropic’s broader safety strategy: users seeking guidance on subjects such as cybersecurity, biology, or chemistry would be redirected to a less capable model, reducing the likelihood that the system could assist with cyberattacks or the development of biological weapons.
A separate provision, however, drew far more scrutiny.
‘I’d Still Have Defended Them… But’
Anthropic had planned to subtly reduce Claude Fable 5’s effectiveness when it detected attempts to use the model for frontier AI research. Because the limitations would not have been disclosed to users, researchers argued the policy amounted to an undisclosed handicap on efforts to leverage Claude in the development of rival models.
Dean Ball, a senior fellow at the Foundation for American Innovation wrote on X, “I’ll be honest that it would have been much more difficult to defend Anthropic against the DoW incursion had that incident occurred after this one. This is the company literally telling their customers, ‘We reserve the right to silently sabotage you.’ I’d still have defended them, because the government trying to destroy a firm is still wrong, but man would it have been a harder case to make.”
Anthropic’s terms of service already prohibit customers from using its systems to train competing AI products, but critics said the hidden restrictions crossed a line by embedding covert performance degradation directly into the model.
Photo: PhotoGranary02/ Shutterstock
