US blocks foreign access to Anthropic’s newest AI models over security risks
Washington has ordered AI firm Anthropic to block all foreign nationals, including its own employees, from accessing its latest models, Fable 5 and Mythos 5, due to national security concerns.
Intelligence analysis by Gemini 2.5 Flash
The US Commerce Department issued an unprecedented export-control measure targeting specific AI models, Fable 5 and Mythos 5, developed by Anthropic. This directive, reportedly prompted by officials finding ways to bypass safeguards, mandates a complete shutdown of these top-tier systems for all foreign access, highlighting Washington's proactive stance on frontier AI regulation.
Imagine a super smart new toy that can do amazing things, but grown-ups worry that if some people play with it the wrong way, it could cause big problems, like sharing secrets. So, the government told the toy maker that only people from their country can use this special toy, even if the toy maker has friends from other countries who helped build it. They want to be super careful with this powerful new toy.
Analysis
The US government has taken a significant step in regulating advanced artificial intelligence by ordering Anthropic, a prominent US-based AI company, to restrict access to its latest models, Fable 5 and Mythos 5, for all foreign nationals. This mandate, issued by the Commerce Department, is notable as it is the first export-control measure specifically targeting AI models.
The directive came shortly after Anthropic's release of Fable 5 and Mythos 5. According to Anthropic's statement, the order extends even to their foreign-born employees, necessitating an abrupt disabling of these models for all their customers to ensure compliance.
Anthropic had previously presented Fable 5 and Mythos 5 as their most powerful models to date. Both models share similar technical foundations. Fable 5 was released to the public with additional safety and security restrictions, which included limitations on discussing sensitive topics like cybersecurity, biology, and chemistry. Mythos 5, conversely, was made available exclusively to trusted partners.
Prior to this, Anthropic had warned about the potential for Fable 5 to be "misused to cause serious damage" if its safeguards were bypassed. The company had also unveiled a preview of the Mythos model to a select group of industry partners under an initiative named Project Glasswing, touting its advanced cybersecurity capabilities. The government's action reportedly followed discoveries by officials of methods to bypass the models' existing safeguards.
Key points
- The US government has ordered Anthropic to block foreign access to its Fable 5 and Mythos 5 AI models.
- This is the first export-control measure to specifically target AI models.
- The directive applies to all foreign nationals, including Anthropic's own foreign-born employees.
- The decision was prompted by national security concerns and reports of bypassed safeguards.
- Fable 5 was publicly released with safeguards, while Mythos 5 was for trusted partners only.
The swift action by the US government could be seen as a strong commitment to national security in the face of rapidly advancing AI, potentially ensuring that powerful models like Fable 5 and Mythos 5 are not misused. This could lead to more robust, nationally aligned AI development that prioritizes safety and controlled access, fostering a more secure technological landscape.
The imposition of such strict export controls on AI models could stifle international collaboration and slow down global AI research and development. It might also create a fragmented AI ecosystem where different regions develop powerful models in isolation, potentially exacerbating geopolitical tensions and hindering the universal benefits of AI advancements.
