On Friday, the US commerce department issued an order restricting the use of Anthropic’s newest models, Fable and Mythos, by non-US nationals, due to national security concerns — prompting the company to suspend its latest tools. There is some irony in the Trump administration, which brags about its hands-off approach to tech, blocking the vocally pro-regulation Anthropic. But the way the move was carried out — giving Anthropic 90 minutes to comply — has raised alarms across the AI industry. Without a clear and consistent approach to policing frontier models, the US risks hampering the development of the world’s most crucial technology.
The order to Anthropic reportedly came after Amazon researchers raised alarms about potential vulnerabilities in Fable that might allow guardrails to be bypassed and users to gain information about security flaws in other software. Anthropic said it had not been given specific details of the US government’s concerns but insisted that a narrow “jailbreak” did not merit recalling a commercial model deployed to hundreds of millions of people. Other systems, including those developed by rivals such as OpenAI, possessed similar capabilities, it added.
The latest cutting-edge models are powerful and the potential national security threats are real. A proper framework to vet the most advanced technologies is needed. The Trump administration recently issued an executive order creating a voluntary framework for US agencies to gain early access to AI models, which, although watered down from a previous version, represented an important first step towards a more coherent approach.