Overview
- On June 12 the Commerce Department’s Bureau of Industry and Security ordered Anthropic to block all foreign‑person access to Fable 5 and Mythos 5, and Anthropic disabled both models globally to comply.
- In Project Glasswing tests run with U.S. agencies Anthropic’s Mythos model identified vulnerabilities in highly sensitive classified systems within hours, though officials say identification does not mean the model exploited those flaws.
- Anthropic told senators that operators linked to Alibaba ran a large-scale distillation campaign that it says used roughly 25,000 fraudulent accounts to generate about 28.8 million exchanges with Claude between April 22 and June 5 to extract model capabilities.
- Cybersecurity leaders warn the shutdown removes powerful defensive tools used to find and patch bugs, while OpenAI’s release of GPT‑5.5‑Cyber with similar benchmark performance has raised questions about why Anthropic faced unique restrictions.
- A U.S. litigation company has sued to overturn the export directive, allied governments are pressing for a vetted 'trusted partner' access framework, and daily technical negotiations between Anthropic and agencies are ongoing.