The White House opposes Anthropic expanding Mythos to 120 companies: they want to use it themselves, but lack enough computing power to share.

According to Beating Monitoring, Anthropic proposed to add approximately 70 companies and organizations to use Mythos, bringing the total to about 120, including the existing roughly 50; the White House refused on the grounds of security and computing power. Mythos was released in early April, capable of detecting and exploiting many critical software vulnerabilities. Currently, it is only being tested by companies managing critical infrastructure, with no plans for public release.

The White House is concerned that Anthropic’s computing power is insufficient. Opening up to 70 more commercial users alone might cause issues for the government using Mythos. White House AI advisor David Sacks publicly stated that the limited release of Anthropic itself indicates that its computing power is inferior to its competitors. Anthropic has signed computing power procurement agreements with Amazon, Google, and Broadcom, but new capacity has not yet come online.

Politically, there has been no easing. The Trump administration criticized Anthropic for hiring several former Biden administration officials and was also dissatisfied with its connections to liberal organizations. One detail that reveals the level of trust: former Anthropic researcher Collin Burns was already scheduled for a government AI model evaluation position, but after the White House leadership found out, they directly replaced him, citing a desire to prevent AI company personnel from sitting in positions that interact with major AI firms.