White House worried about compute limits as it blocks wider access to Anthropic's Mythos

The White House has rejected Anthropic's plan to expand access to its AI model Mythos to roughly 70 additional companies, the Wall Street Journal reports. Mythos can find and exploit software vulnerabilities, which raises significant security concerns. Currently, around 50 organizations have access through Project Glasswing, including critical infrastructure operators and government agencies such as the NSA.

AI advisor David Sacks and other government officials believe Anthropic has less compute capacity than its competitors, according to the report. Despite new deals with Amazon, Google, and Broadcom, it will take time before additional capacity comes online. The administration is concerned that adding more users could limit the government's own access to the model.

At the same time, the White House is looking for ways to continue working with Anthropic after the Pentagon classified the company as a supply chain risk.