Anthropic announced Friday that it has blocked every customer from using its Fable 5 and Mythos 5 artificial‑intelligence models. The decision follows a government order received on June 12 that required the company to suspend access for all foreign nationals, regardless of location or employment status, on grounds of national‑security risk.
According to the company’s statement, the directive does not affect its other offerings, including the Claude chatbot and earlier models. Anthropic framed the action as a compliance measure, noting that the order specifically targets the two newest models launched just days earlier.
Why the government intervened
Officials have not disclosed the precise security concerns behind the order. Anthropic, however, believes the impetus was a reported method for jailbreaking Fable 5. The company said it received verbal evidence of a narrow, non‑universal jailbreak that an unnamed entity shared with authorities. While the details remain classified, Anthropic indicated that the potential exploit could allow users to bypass safeguards designed to prevent misuse in cybersecurity contexts.
Fable 5, introduced on June 9, was marketed as a public‑facing version of Anthropic’s high‑end Mythos model. Mythos, described as a state‑of‑the‑art cybersecurity system, is currently limited to Anthropic’s Project Glasswing partners. In internal tests, Fable 5 demonstrated capabilities that surpassed earlier releases, even beating the video game Pokémon FireRed—a benchmark Claude could not reach.
Anthropic highlighted the extensive protective measures baked into Fable 5. The company said it implemented “strong safeguards” to lower the likelihood of the model being used for malicious cybersecurity tasks. Some users have complained that these safeguards are overly broad, but Anthropic defended them as necessary to limit abuse.
The firm also acknowledged that no AI model can be entirely immune to jailbreak attempts. Its strategy, it explained, is to make any successful exploit either narrowly targeted or prohibitively costly to develop. Continuous monitoring is intended to detect and shut down attacks swiftly.
Anthropic’s response and next steps
While complying with the government order, Anthropic expressed disagreement with the notion that a single potential jailbreak warrants recalling a commercial model. The company pledged to share further details within the next 24 hours, asserting that any action to block unsafe deployments should follow a transparent, fair, and technically grounded statutory process.
“We have consistently called for oversight that is clear and based on technical facts,” the statement read. “This action does not adhere to those principles.”
For now, the suspension remains in place, and customers awaiting access to Fable 5 and Mythos 5 must await further clarification from both Anthropic and government regulators.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.