Trump Administration Blocks Anthropic’s Mythos Rollout
The Trump administration has told Anthropic it disagrees with a controlled rollout that would bring its advanced cybersecurity AI, Mythos, to around 70 additional companies, citing concerns over compute power and security. According to Bloomberg, which cited an anonymous administration official, the White House opposes the expansion of access to this powerful AI tool.
April 30, 2026 - 8:52 am
This development follows a report by the Wall Street Journal and was first announced through Anthropic’s Project Glasswing initiative in early April. Mythos is an AI model capable of autonomously identifying and exploiting vulnerabilities in critical software, which Anthropic considers too dangerous for general release.
Instead of a broader rollout, Anthropic has been allowing a limited set of organizations to test Mythos on their systems. However, the Trump administration's objection comes down to two main concerns:
- Security Risks: The potential for misuse and the impact on government access to the model.
- Compute Power: Concerns that Anthropic lacks sufficient computing power to serve 120 entities without compromising the government’s use of Mythos.
The compute issue is significant as part of Anthropic's current $900 billion fundraising effort is aimed at securing the necessary infrastructure to run Mythos at scale. Agencies like the National Security Agency (NSA) are already using Mythos, and the White House's objection seems to stem from a desire to control access to this powerful tool.
A recent security breach, where unauthorized users gained access to Mythos on a private online forum, further complicates matters. This incident highlighted the challenges of containing access to such advanced AI models and intensified government apprehension about expanding the user base.
Despite these objections, the capabilities of Mythos are undeniable. The model autonomously discovered thousands of zero-day vulnerabilities across operating systems and web browsers in testing and successfully completed complex cybersecurity tasks. These risks are not hypothetical; they are real concerns that have prompted both security warnings and a desire to control access through the NSA's involvement.