Anthropic's Trip to Washington for Negotiations on Lifting the Ban Ends in Failure, Claude Fable 5 Still Maintains Global Ban

According to Sentinel Beating monitoring, the U.S. Department of Commerce held an emergency working meeting with Anthropic on Monday, June 15, regarding the recent AI model export control. However, the talks ultimately ended in a stalemate, with Claude Fable 5 still under a global ban. The previous U.S. government's control order was primarily motivated by security concerns, fearing that once the security defenses of Fable 5 are bypassed, it could degrade and unleash potent capabilities such as Mythos-level powerful network attacks and vulnerability exploits, which could be weaponized by other countries' military intelligence agencies.

This closed-door meeting in Washington was led by Anthropic's co-founder and Chief Compute Officer Tom Brown, Frontiers Red Team lead Logan Graham, and senior security researcher Nicholas Carlini. Officials from the National Cybersecurity Directorate and the Department of Commerce were present to listen to the discussions. The team reiterated that the U.S. government's concerns about the jailbreaking risk of Fable 5 were exaggerated, but officials from the Department of Commerce and the Center for AI Standards and Innovation (CAISI) researchers remained unconvinced. The government side insisted that Anthropic must thoroughly address the jailbreaking vulnerability.

Currently, both parties are urgently discussing the next steps for a resolution. U.S. Secretary of Commerce Howard Lutnick remotely joined the meeting from the G7 Summit in Évian-les-Bains, France, and maintained regular calls with Anthropic's CEO Dario Amodei. As both are attending the summit in France, Lutnick and Amodei are expected to engage in face-to-face discussions at the G7 Summit site this week. In response to the ban, more than 80 cybersecurity experts, including executives from companies like NVIDIA, Adobe, and Zoom, as well as academic scholars, jointly signed an open letter on June 14 in support of Anthropic, stating that the ban would not only harm the cybersecurity posture but also weaken the U.S.'s leadership position in the AI field.