A coalition of cybersecurity professionals is urging the Trump administration to reconsider its restrictions on Anthropic’s advanced AI models. The group argues that these limitations could inadvertently benefit U.S. adversaries instead of enhancing national security.
Anthropic’s AI Models Taken Offline
As of Friday, Anthropic has removed its cutting-edge AI models, Fable 5 and Mythos 5, from use to adhere to governmental directives. The San Francisco-based AI firm contends that the government’s measures are unnecessary, refuting the supposed security threats that prompted these actions.
The company has restricted access to its latest technologies, claiming their potential to outperform human cybersecurity experts in identifying and exploiting computer vulnerabilities. Previously, Anthropic engaged in discussions with the White House regarding the capabilities of these models.
Community Calls for Transparency
In a letter sent Sunday, over 100 cybersecurity authorities, including representatives from Adobe and Nvidia, petitioned the U.S. government to revoke the export control directives on Anthropic’s AI models. The signatories advocate for a transparent and scientific approach to AI risk assessments moving forward.
The letter highlights that while Anthropic’s Mythos models are proficient in identifying software vulnerabilities, they are not exceptional in this domain. Many experts routinely employ alternative foundational and open-source models for security evaluations and training.
Concerns Over Global AI Advancements
Signatories of the letter emphasize the danger of removing top-tier cyber defense tools without substantial justification, especially as international adversaries advance rapidly. The letter warns that China’s AI models are only marginally behind American counterparts, with the potential for undisclosed government capabilities.
The U.S. government’s export controls represent a significant measure to limit access to sophisticated AI models. Although Anthropic released a less advanced version of Fable last week, access to the more sophisticated Mythos remains restricted due to security concerns.
Future Implications and Ongoing Disputes
The Commerce Department has yet to comment on the situation, which follows President Trump’s executive order establishing a framework to assess national security risks associated with advanced AI systems. This framework allows AI developers to voluntarily participate in a month-long evaluation prior to public release.
Tensions persist between the Trump administration and Anthropic, which aims to implement safety measures in AI development to optimize its benefits for economic and national security. A contract disagreement with the Pentagon has led to Anthropic challenging a potential designation as a supply chain risk, a move it disputes in federal courts.
