AI company Anthropic announced on Friday that it has removed its latest artificial intelligence models, called Fable 5 and Mythos 5, from public access in compliance with a directive from the Trump administration aimed at preventing their use by individuals from other countries.
This move represents a significant effort by the U.S. government to limit the availability of highly advanced AI models. While Anthropic recently made Fable available to the public, it has tightly controlled access to the even more advanced Mythos due to concerns about cybersecurity.
In a statement, Anthropic expressed disagreement with the government’s approach, noting that the directive received on Friday did not specify the national security issues prompting the action. The company emphasized the importance of a transparent and fair process based on technical facts for blocking unsafe deployments.
Anthropic referred to the situation as a “misunderstanding” and expressed hope for the swift restoration of access to the models. The Commerce Department did not provide an immediate response to requests for comment.
This development comes following President Donald Trump’s executive order, issued 10 days earlier, establishing a framework for evaluating national security risks posed by advanced AI systems before their public release. Anthropic had previously highlighted the potential risks associated with Mythos, particularly in terms of software vulnerabilities that could impact economies, public safety, and national security if exploited.
Bank of Canada governor Tiff Macklem has also weighed in on the discussion around Mythos, emphasizing the need for proactive planning to address the challenges posed by rapidly evolving AI technologies. Macklem stressed the importance of robust cybersecurity measures to safeguard against vulnerabilities exposed by these advanced AI models.
The ongoing dialogue underscores the need for a comprehensive approach involving companies, regulators, and policymakers to navigate the complexities of AI development and deployment effectively.
