Anthropic shuts down most powerful AI models due to security warnings

The US government ordered Anthropic on Friday to immediately suspend access to its two most powerful AI models, Claude Fable 5 and Claude Mythos 5, citing national security concerns. Anthropic stated on its X page that it complied with the order but believes the government's decision is misguided. The directive requires blocking these models not only for foreign nationals but for all users worldwide. This is reported by Techcrunch.com reports.

The Mythos model is Anthropic's most advanced development, possessing extraordinary capabilities in identifying software vulnerabilities. For this reason, the company did not release it to the public, providing it only to 50 selected organizations such as Amazon, Apple, Google, Microsoft, and CrowdStrike for cybersecurity protection. Fable 5 was a public version based on Mythos, but with restrictions in sensitive areas like cybersecurity and biology.

The government's decision is reportedly due to the discovery of a jailbreak vulnerability in the Fable 5 model's security system. Anthropic argues that such vulnerabilities exist in other public models like GPT-5.5 from OpenAI and that this is common for cybersecurity experts. The company claims its security systems operate via independent classifiers and that no real dangerous content has been generated.

Anthropic representatives did not hide their dissatisfaction with the government's action. They believe that if such strict standards are applied to the entire industry, it could halt the launch of all new AI models. This is an unexpected blow for the company, which is planning an IPO this year.

Observers note that Anthropic's marketing of its models as extremely dangerous has ultimately backfired. OpenAI CEO Sam Altman previously called this approach by Anthropic "fear-based marketing." Now, this very caution has led to strict government oversight and restrictions on the company's business operations.