Anthropic disables access to top-tier AI models after US ban on foreign use
AI company Anthropic has disabled access to its Fable 5 and Mythos 5 models for all users following a US government order citing national security concerns, related to a potential "jailbreak" vulnerability.
Intelligence analysis by Gemini 2.5 Flash
Anthropic, an American AI startup, has complied with a US government directive to suspend foreign access to its most powerful AI models, Fable 5 and Mythos 5, due to undisclosed national security concerns over a potential vulnerability that could bypass safeguards. The company disputes the necessity of the broad ban, arguing it could halt future AI model deployments across the industry.
Imagine a super smart computer program called Fable 5 that can help with lots of things. The US government told the company that made it, Anthropic, to turn it off for everyone outside the US, and even for some people inside, because they're worried someone might find a secret way to make it do things it shouldn't, like finding weaknesses in other computer programs. Anthropic thinks it's a small problem, but the government is taking it very seriously to keep everyone safe.
Analysis
Anthropic, a leading AI company, announced on Friday its decision to "abruptly disable" its top-tier AI models, Fable 5 and Mythos 5, for all users worldwide. This action comes in response to an export control directive issued by the US government. The directive mandated the suspension of access to these models for foreign nationals, citing unspecified national security concerns. Anthropic stated that it was not provided with specific details regarding the nature of these concerns.
Alleged Vulnerability
The company believes the government's concern stems from a potential "jailbreak" method that could bypass safeguards in Fable 5. This vulnerability, if exploited, might allow the model to be used in identifying software vulnerabilities, an area where its use is typically restricted. Anthropic, however, disagrees with the government's assessment, calling the alleged jailbreak "narrow, non-universal" and questioning whether it warrants recalling a commercial model deployed to hundreds of millions of people.
Growing Tensions and Policy Shift
This incident underscores the increasing friction between AI developers and government regulators over how to evaluate and manage the risks associated with AI models, particularly concerning methods like "jailbreaks" that can circumvent safety protocols. The move also signifies a significant shift in US policy, which traditionally focused export controls on AI chips and tools rather than direct restrictions on foreign access to AI models themselves. This directive marks an escalation of US efforts to curb foreign adversaries' AI capabilities.
Past Disputes and Industry Impact
Anthropic's relationship with the US government has been strained previously, notably after the company refused to permit the US military to utilize its AI models for domestic surveillance and fully autonomous weapons systems. This refusal led to Anthropic being placed on a supply chain blacklist. Kirsten Davies, the Pentagon's chief information officer, voiced support for prioritizing national security, stating that "Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation. America First. Always." Anthropic is currently preparing for a US IPO, placing it ahead of rivals like OpenAI in seeking public market access. The company argues that if this strict standard were applied industry-wide, it would effectively halt the deployment of all new frontier AI models.
Cybersecurity Risks
Experts have raised concerns that if Mythos models were to fall into the wrong hands, they could significantly accelerate sophisticated cyberattacks, especially targeting critical sectors like banking, which often rely on complex and aging technological infrastructures. A US official confirmed that the Commerce Department had indeed issued the export control directive.
Key points
- Anthropic disabled access to its Fable 5 and Mythos 5 AI models for all users globally.
- The action follows a US government export control directive citing unspecified national security concerns.
- The government believes a "jailbreak" vulnerability could allow Fable 5 to be used for identifying software vulnerabilities.
- Anthropic disputes the severity of the alleged vulnerability, calling it a "narrow, non-universal jailbreak".
- This incident highlights growing tensions between AI developers and regulators over assessing AI risks and export controls.
- Experts warn that Mythos models in the wrong hands could accelerate sophisticated cyberattacks.
- Anthropic is working to restore access, arguing that such a broad ban could halt future AI model deployments across the industry.
Despite the immediate access restrictions, Anthropic is actively working with the US government to address the "misunderstanding" and hopes to restore access to its advanced models soon. If a clear and fair regulatory framework emerges from these discussions, it could establish a balanced approach to AI safety while allowing continued innovation and global access to responsible AI technologies.
The broad suspension of Anthropic's top-tier AI models, driven by a potentially minor "jailbreak" concern, could set a restrictive precedent for the entire AI industry. This stringent approach might significantly slow down or even halt the deployment of new, advanced AI models from all providers, stifling innovation and limiting global access to cutting-edge AI capabilities due to overzealous regulation.
