Anthropic said on Friday it would “suddenly disable” its most advanced artificial intelligence (AI) models. cloud fable 5 And mythos 5for all users after the US government ordered the suspension of access to the models for foreign nationals, inside or outside the US, citing national security concerns.
The AI company said it received an order at 5:21 pm ET directing it to suspend all access to the models by foreign nationals. It said it believed there was a “misunderstanding” and was working to restore access to the models as soon as possible. The export control directive will not affect access to other models.
“Our understanding is that the government believes it has become aware of a method to bypass or ‘jailbreak’ Fable 5,” the company said.
“We reviewed the performance of this specific technique being used to identify a small number of previously known vulnerabilities. All of these vulnerabilities appear relatively simple, and we have found that other publicly available models are also able to find them without the need for bypass.”
This unexpected move comes just days after the launch of CloudFable 5 and its counterpart Mythos 5, which uses the same underlying model but with security measures removed in some areas such as cybersecurity. The latter, which has been described as having “the strongest cybersecurity capabilities of any model in the world”, remains accessible to a verified group of cyber defenders and critical infrastructure operators.
Anthropic stressed that it has implemented “strong” guardrails to prevent misuse of its models for cybersecurity-related tasks. Specifically, it belongs to a set of security classifiers that are used to detect potential abuse, including jailbreak attempts, and prevent the core model from responding.
Cybersecurity classifiers are designed to block harmful single-turn requests related to cyberattack planning, exploit development, or defense evasion, with the company noting that Mythos-class models are efficient at finding and exploiting software vulnerabilities, giving attackers a strategic advantage.
Last week, Anthropic revealed that its Mythos-class model could remediate newly disclosed software vulnerabilities in hours or, in some cases, minutes instead of weeks, converting n-days into n-hours. The findings suggest that Frontier models may be just as good at rapidly weaponizing publicly disclosed flaws.
Anthropic’s Red Team said, “A single operator can now turn a month’s worth of patches into a working exploit in a single afternoon – for a few thousand dollars and without any special expertise.” “This means that the typical patching playbook that software developers use today – with monthly release cadences, multi-week phased rollouts, and lag between pre-release and stable channels – is no longer valid.”
The security of the Fable 5 means that questions on cybersecurity topics will get a response from the company’s next capable model, Cloud Opus 4.8.
In its latest statement, the company argued that no universal jailbreak method has been developed against the latest models to date, and said that third-party and internal red-teaming exercises have found its security measures to be “significantly more effective than any previously deployed models.”
Furthermore, Anthropic claimed that “complete jailbreak resistance” is not possible for any model provider, as every security measure used by the industry is susceptible to non-universal jailbreaks that are “effective in very limited contexts or require additional effort to adapt to each new situation.”
“To date, the government has given us only verbal evidence of a potentially narrow, non-universal jailbreak, which essentially involves asking the model to read a specific codebase and fix any software flaws.”
“Our understanding is that a potential jailbreak was shared with the government. We have reviewed a report, which we believe to be the basis for the government directive, and have validated that the level of capability demonstrated there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by defenders securing systems.”
Anthropic also pointed out that it is the government’s job to prevent unsafe AI deployments, but it said the discovery of a “narrow potential jailbreak” should not be a reason to recall widely deployed commercial models. It states that the legislative process should be “transparent, fair, clear and based on technical facts”.
Earlier this year, the US Defense Department branded Anthropic a “supply chain risk” as the cloud-maker sought to draw red lines on military use of its technology. The company has filed two lawsuits to block the designation.
