|
Anthropic said there is a "misunderstanding" behind the US government directive that forced it to suspend Fable 5 and Mythos 5, and the company is working to restore access as soon as possible. The AI firm also warned that if the government applies its current standard to narrow jailbreak findings, it would "essentially halt all new model deployments for all frontier model providers."
The company announced the suspension on Friday after receiving an export control directive citing national security authorities. The order requires Anthropic to block all access to Fable 5 and Mythos 5 by any foreign national, "whether inside or outside the United States, including foreign national Anthropic employees." Access to all other Anthropic models remains unaffected. Anthropic said it disagrees with the basis of the directive but is complying while engaging with officials. "We are complying with the government's legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic stated. The company said the government cited a potential jailbreak but provided no specific details. "Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5," Anthropic said. After reviewing a demonstration, it found the technique surfaced "a small number of previously known, minor vulnerabilities" that "other publicly-available models are able to discover... without requiring a bypass." Anthropic defended Fable 5's safeguards and testing. "We have instituted strong safeguards that greatly reduce the likelihood that Fable is misused for tasks related to cybersecurity... In fact, our safeguards are so strong that many users have complained that they are overly broad," the company said. It added that pre-launch red-teaming with the US government, UK AISI, multiple third-party organizations and internal teams ran "for thousands of hours in total" and showed Fable's safeguards are "substantially more effective than those of any previously deployed model." It also noted "no testers have yet been able to find a universal jailbreak." On the specific concern, Anthropic said the capability cited is widely available. "To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws," Anthropic said. "We validated that the level of capability displayed there is widely available from other models including OpenAI's GPT-5.5, and is used every day by the defenders who keep systems safe." The company said it supports government authority to block unsafe deployments but wants a clearer process. "As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles," Anthropic said. It added it will share more details over the next 24 hours while working to restore access. (ANI)
|
