- Anthropic and Trump administration officials held senior-level talks Monday — led by the Commerce Department and National Cyber Director Sean Cairncross — seeking to resolve the security dispute that forced Fable 5 and Mythos 5 offline.
- After the White House restricted foreign access to Fable 5, Anthropic shut down both models for all users to comply — an unprecedented move for a leading US AI company.
- The dispute was triggered when Amazon researchers found a way to bypass Fable 5’s guardrails days after launch; Amazon CEO Andy Jassy alerted administration officials including Treasury Secretary Scott Bessent.
- Anthropic, valued at $965 billion and heading toward an IPO, broke its silence Monday: “Both parties are working quickly to get this resolved.”
What Happened?
Anthropic and senior Trump administration officials continued negotiations Monday over the security concerns that led to an unprecedented shutdown of the company’s two most powerful AI models. After the White House restricted foreign access to Fable 5 — a general-purpose model capable of potent cyberattacks — Anthropic pulled both Fable 5 and the more powerful Mythos 5 entirely, cutting off all users globally. Monday’s talks included top Anthropic security researchers Nicholas Carlini and Logan Graham, and were led by the Commerce Department and National Cyber Director Sean Cairncross. Commerce Secretary Howard Lutnick participated in weekend discussions. In its first public statement since the shutdown, Anthropic said “both parties are working quickly to get this resolved,” framing the talks as part of a shared commitment to protecting US critical infrastructure.
Why It Matters?
The episode is a collision between two things Anthropic has staked its identity on: responsible AI development and a cooperative relationship with government. The shutdown was triggered by Amazon researchers who discovered a way around Fable 5’s safety guardrails shortly after its release — and Amazon CEO Andy Jassy took that finding directly to administration officials. Amazon is both Anthropic’s biggest investor and a major government cloud partner, making its role in triggering the crackdown particularly pointed. With Anthropic valued at $965 billion and an IPO on the horizon, every day the models remain offline compounds financial and reputational pressure. Some administration officials want an acknowledgment from Anthropic that its Fable rollout and government communications could have been handled better — a concession Anthropic has so far declined to make explicitly.
What’s Next?
Both sides have strong incentives to reach a deal quickly. The administration wants to show it can responsibly oversee AI without stifling a sector where US leadership is a stated priority. Anthropic needs its flagship models back online before competitors gain ground. The resolution will likely involve some combination of enhanced access controls for foreign users and a clearer pre-release evaluation process with the Commerce Department unit that already reviewed Fable before launch. How the dispute is resolved will set a precedent for how the US government manages security reviews of future AI model releases.
Source: The Wall Street Journal
