BotBeat
...
← Back

> ▌

AnthropicAnthropic
POLICY & REGULATIONAnthropic2026-06-16

Trump Administration Halts Anthropic's Claude Mythos 5 and Fable 5 Over Export Control Concerns

Key Takeaways

  • ▸US government issued export control directive forcing Anthropic to disable Claude Mythos 5 and Fable 5 immediately, affecting foreign national access including company employees
  • ▸Anthropic received 90-minute ultimatum on Friday and escalated to C-suite negotiations with Treasury Secretary, Commerce Secretary, and National Cyber Director
  • ▸Government concerned about jailbreak vulnerability and potential access by China-linked entities to advanced AI models
Sources:
Hacker Newshttps://www.theverge.com/ai-artificial-intelligence/950412/anthropic-trump-adminstration-claude-mythos-fable-5-export-controls↗
Hacker Newshttps://www.theatlantic.com/technology/2026/06/trump-anthropic-export-control-ai-race/687555/↗
Hacker Newshttps://www.wired.com/story/anthropic-is-still-at-odds-with-the-white-house-over-claude-fable-5/↗

Summary

The Trump administration issued an export control directive on Friday requiring Anthropic to immediately suspend access to its Claude Mythos 5 and Fable 5 models by foreign nationals, both inside and outside the US. The company received a 90-minute ultimatum and was forced to completely disable the products it had been promoting, with CEO Dario Amodei spending the weekend negotiating directly with US Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross.

The government's action reportedly stemmed from concerns about a potential jailbreak vulnerability in Fable 5 and fears that China-linked entities might gain access to the technology. Sources indicate the administration called Anthropic at 1 PM ET with the ultimatum, citing a report of bypass methods that could compromise the models' safety guardrails.

AnthropIc disputed the government's characterization of the threat, releasing a statement arguing that the reported vulnerability was "narrow and non-universal" and that similar capabilities were already widely available in competitors' models, including OpenAI's GPT-5.5. The company emphasized that it had shared the jailbreak report with the government voluntarily.

  • Anthropic disputes severity of threat, arguing reported vulnerability is narrow and non-unique compared to competitors like OpenAI
Large Language Models (LLMs)Generative AICybersecurityGovernment & DefenseRegulation & PolicyAI Safety & Alignment

More from Anthropic

AnthropicAnthropic
POLICY & REGULATION

Global Nobel Laureates Issue Rome Declaration Calling for Coordinated AI Slowdown and Safety Measures

2026-08-02
AnthropicAnthropic
POLICY & REGULATION

Australian Booksellers Caught in AI's Destructive Data-Harvesting Supply Chain

2026-08-01
AnthropicAnthropic
RESEARCH

IssueTrojanBench Security Study Reveals Critical Vulnerabilities in AI Coding Agents

2026-08-01

Comments

Suggested

Hugging FaceHugging Face
OPEN SOURCE

Strangers Pretrain 15M-Parameter Language Model Using GitHub Actions and Hugging Face PRs

2026-08-02
General AI ResearchGeneral AI Research
RESEARCH

Research Identifies Fundamental Trilemma: LLM Safeguards Cannot Simultaneously Provide Reliable Safety, Useful Capability, and Open Access

2026-08-02
Alibaba (Cloud)Alibaba (Cloud)
INDUSTRY REPORT

Token Diplomacy: China Positions Open-Source AI as Global Strategic Resource

2026-08-02
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us