The US government has partially lifted restrictions on Anthropic's Claude Mythos 5 AI model, allowing the company to provide access to over 100 trusted US organizations. This decision comes after Anthropic worked with the government to address concerns around the model's risks. In a letter to Anthropic's cofounder Tom Brown, US Commerce Secretary Howard Lutnick stated that the company has implemented sufficient safeguards, enabling the model's release to a select group of cyber defenders and infrastructure providers. However, the government has not yet permitted a broader rollout of the model or commented on the status of Claude Fable 5, the consumer-facing version. Anthropic is now working to restore access to Mythos 5 for the approved organizations, which may include foreign national employees. This development marks a significant step forward for Anthropic, but raises questions about the US government's approach to regulating AI model releases. The company's experiences have highlighted the need for developers to obtain explicit government approval before releasing advanced AI models. The partial reinstatement of Mythos 5 follows a series of events, including an export control directive and concerns raised by Amazon and the National Security Agency. Anthropic's interactions with the government have been costly, with the company suing the Trump administration earlier this year over a supply chain risk designation. As the situation continues to unfold, it remains to be seen how the US government will balance the need to regulate AI with the need to support innovation in the field.

Source: Wired