White House Blocks Anthropic Plan to Expand Mythos AI Access

The White House blocked a proposal by Anthropic PBC to grant approximately 70 additional companies and organizations access to its Mythos AI model. Administration officials told reporters that the model possesses the capability to detect and exploit critical software vulnerabilities, which could facilitate dangerous cyberattacks and jeopardize national security.

Beyond security risks, the administration expressed concerns regarding computing resources. Officials argued that Anthropic lacks the sufficient power to support an expanded user base without compromising the government's own capacity to utilize the technology effectively. These restrictions aim to ensure that federal access to the model remains stable and prioritized over private sector expansion.

This conflict occurs within a broader period of strain between the Trump administration and Anthropic. The relationship is marked by ongoing legal disputes with the Pentagon and administration criticism of the company's ties to former officials from the Biden administration. These tensions recently manifested in the administration's decision to block the hiring of Anthropic researcher Collin Burns for a government AI evaluation role, signaling a hardening stance against the company's influence in federal AI policy.