Anthropic halts all services for top AI models Claude Fable 5 and Claude Mythos 5

[Financial News]  


The AI startup Anthropic said on the 12th, local time, that it had fully suspended services for its top AI models, Claude Fable 5 and Claude Mythos 5. The move followed the U.S. government's announcement that it would restrict foreign access for national security reasons.
The Trump administration's policy is to treat foreign access to advanced AI models as an export issue, much like controls on exports of advanced AI semiconductors. Foreign nationals in the United States are also barred from using the models.
According to CNBC and the Financial Times, Anthropic said, "The U.S. government issued export control guidance under national security directives that completely blocks all foreign nationals from accessing Claude Fable 5 and Claude Mythos 5." Anthropic said the order came at 5:21 p.m. Eastern Time.
The full suspension came just days after Anthropic made the two powerful AI models available to several industry-leading companies and countries. Anthropic suggested that the government may still believe unresolved security concerns remain, even after it released Claude Fable 5 for public use by fixing security flaws in Claude Mythos, which was unveiled in April and had drawn warnings from the government and financial institutions over serious security risks.
Anthropic said the government appears to believe that Claude Fable 5's safeguards could be bypassed through so-called "jailbreaking," allowing the AI to be misused for hacking or the production of biological weapons. The company said that view appears to be a misunderstanding.
Anthropic pointed out that the jailbreaking methods the government appears to have identified are widely used with other AI models and are employed daily by experts responsible for system security.
It then pushed back, saying, "The government should have the authority to block the deployment of unsafe AI through transparent, fair, and technically grounded legal procedures, but this measure does not meet those principles."
The company appeared to suggest that it could pursue legal action if the issue is not resolved.
Anthropic is already engaged in a legal battle against the Trump administration's decision to place it on the sanctions list.
Earlier, the U.S. Department of Defense labeled the AI startup a threat to national security, saying it posed risks to the supply chain after Anthropic opposed using its AI models in warfare. The practice of placing foreign companies such as Huawei on the sanctions list over alleged threats to U.S. national security has now expanded to U.S. firms.
Because of the Defense Department's sanctions listing, defense contractors must prove that they are not using Anthropic's AI models in order to take part in Pentagon procurement projects.

dympna@fnnews.com Song Kyung-jae Reporter