Key Points
- Andy Jassy, Amazon’s CEO, alerted Trump administration leaders to cybersecurity vulnerabilities in Anthropic’s Fable 5 after internal testing revealed exploit risks
- Federal officials demanded Anthropic address security flaws or remove the model from service; President Trump authorized complete international access restrictions
- In response to export control requirements, Anthropic deactivated both Fable 5 and Mythos models for its entire user base
- The company characterized the security issues as “relatively basic” and noted comparable features are available in other publicly accessible AI systems
- The restrictions may negatively impact Anthropic’s planned initial public offering while strengthening competitor positions like OpenAI
Amazon’s chief executive, Andy Jassy, engaged in direct communications with federal authorities, including Treasury Secretary Scott Bessent, regarding security vulnerabilities discovered in Anthropic’s Fable 5 artificial intelligence system. Internal testing by Amazon’s research team utilized targeted queries that prompted the model to produce data potentially useful for executing cyber intrusions.
Such outputs should have been prevented by the model’s integrated safety mechanisms. After Amazon shared these discoveries with White House personnel and security agencies, high-level consultations were convened to determine appropriate action.
Federal authorities presented Anthropic with an ultimatum: remediate the vulnerabilities or suspend model operations. During Friday discussions between Anthropic’s chief executive Dario Amodei and government representatives, certain officials perceived reluctance from the company to collaborate with federal security specialists on developing solutions.
Administration decision-makers determined that imposing foreign access limitations represented the most effective approach to mitigating potential threats. President Trump authorized the restriction despite expressing concerns that such measures might impede artificial intelligence advancement.
Company Disables Flagship AI Systems
After receiving the directive, Anthropic terminated access to both Fable 5 and its Mythos system across its entire user population—extending beyond just international users—to maintain full compliance with newly implemented export regulations.
The organization described the identified vulnerabilities as fundamentally unsophisticated. Company representatives emphasized that alternative AI platforms available to the public possess similar information-retrieval capabilities.
Andrew Morris, a cybersecurity expert and founder of GreyNoise Intelligence, examined Amazon’s research results. His assessment confirmed that while Fable 5 demonstrated ability to detect software vulnerabilities across at least four different applications, investigators found no indication that researchers successfully accessed the more hazardous functionality of transforming identified bugs into operational attack code.
Anthropic has consistently promoted safety protocols as a fundamental organizational principle. The company previously postponed wider distribution of Mythos following White House guidance and maintains ongoing collaboration with a federal AI evaluation division prior to launching new systems.
Implications for the Company’s Future
The situation presents significant challenges for Anthropic’s business trajectory. The organization is currently advancing preparations for a possible initial public offering potentially scheduled for autumn 2025.
With its premier models unavailable, existing clients may migrate to alternative providers. OpenAI has developed its own cybersecurity-oriented model and maintains active dialogue with Trump administration representatives.
Friction between Anthropic and federal authorities predates this incident. The Pentagon previously classified Anthropic as a security concern, a determination the company is contesting through two ongoing legal proceedings.
National Cyber Director Sean Cairncross and Commerce Secretary Howard Lutnick participated in deliberations preceding the access ban. The Commerce Department maintains jurisdiction over export control policies affecting critical technologies.
Commerce’s prohibition currently blocks foreign governmental entities, commercial organizations, and private individuals from utilizing Fable and Mythos. A significant portion of Anthropic’s research personnel are international nationals, which the company indicated effectively prevents their continued involvement with these models.
White House artificial intelligence advisor David Sacks characterized the restriction as implemented “reluctantly” and conveyed optimism that Anthropic will resolve the security concerns, enabling eventual restoration of general model availability.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
