Loading IndicatorLoading Indicator

US Orders Anthropic to Halt Fable 5, Mythos 5 for Foreign Users on Security Concerns

Source
Doohyun Hwang

Summary

  • The US government said it had imposed export controls on Anthropic’s latest AI models, Fable 5 and Mythos 5, on national security grounds, triggering an immediate global suspension of the services.
  • The order applies equally to foreign nationals inside and outside the US, as well as Anthropic’s foreign employees, while Anthropic’s other AI models will continue operating normally.
  • Anthropic pushed back, saying concerns over jailbreak techniques are excessive, and described the episode as stemming from a misunderstanding with the government. The company said it would work closely with the government to restore service quickly.

Forecast Trend Report by Period

Loading IndicatorLoading Indicator
Photo: Shutterstock
Photo: Shutterstock

The US government has ordered export controls on Anthropic’s latest AI models, Fable 5 and Mythos 5, citing national security concerns. Anthropic immediately suspended the services for customers worldwide.

Anthropic announced on June 12 that the government had directed it to block all foreign access to Fable 5 and Mythos 5. The order applies to foreign nationals both inside and outside the US, as well as foreign employees at Anthropic. Service for the two models has been halted, while the company’s other AI models continue to operate normally.

Official US documents indicate the government objected to the discovery of so-called jailbreak techniques that can disable Fable 5’s safety guardrails. Anthropic was also presented with verbal evidence that the model could bypass cybersecurity systems through certain prompts.

Anthropic expressed regret over the move and strongly disputed the government’s reasoning. The company said the cited bypass technique amounts to little more than reading a specific codebase and fixing software bugs. It added that comparable capabilities are already widely used every day for defensive security purposes in other publicly available models, including OpenAI’s GPT-5.5.

The company also argued that completely preventing jailbreaks is impossible with current technology. No method has been found to broadly disable its safety systems, Anthropic said. It called it excessive to fully shut down a commercial model used by hundreds of millions of people over the possibility of simple, limited workarounds.

Anthropic added that applying the same standard across the industry would effectively halt new model launches by all advanced AI companies. The company said the government’s blocking order does not meet the principle that such measures should be transparent, fair and based on technical facts. It also described the episode as stemming from a misunderstanding with the government and said it would work closely with authorities to resume service as soon as possible.

Doohyun Hwang

Doohyun Hwang

cow5361@bloomingbit.ioKEEP CALM AND HODL🍀
hot_people_entry_banner in news detail bottom articleshot_people_entry_banner in news detail mobile bottom articles

What do you think about this news?








PiCK News