The U.S. government is reportedly exploring limited access to a powerful artificial intelligence system developed by Anthropic, as concerns grow over both its cybersecurity potential and associated risks.
According to reports from Bloomberg, the company’s advanced model—known as Mythos—is being considered for deployment across select federal agencies under a controlled initiative called Project Glasswing.
A High-Stakes AI Deployment
The initiative would allow certain organizations to use a restricted version of the unreleased Claude Mythos for defensive cybersecurity purposes.
Early reports suggest the model has already identified thousands of vulnerabilities across operating systems, browsers, and other widely used software systems, highlighting its powerful code analysis capabilities.
However, experts caution that the same capabilities that make the system valuable for defense could also be used to generate offensive cyber strategies if misused.
Government Coordination and Safeguards
According to Bloomberg, Gregory Barbaccia, who serves as the federal chief information officer at the White House Office of Management and Budget, informed government departments that protections are being developed before any rollout.
In an internal email, he emphasized collaboration with AI providers, industry partners, and intelligence agencies to ensure proper safeguards are in place before broader access is granted.
The communication, reportedly titled “Mythos Model Access,” did not confirm a final rollout timeline or specify how agencies would use the system.
Strategic AI Competition
The development comes amid ongoing tensions and competition between major AI firms and government agencies over advanced model access and security applications.
Anthropic has been actively working with U.S. officials on controlled AI deployment strategies, even as broader industry relationships shift and evolve.
Co-founder Jack Clark has previously noted discussions with U.S. policymakers regarding the responsible use of advanced AI systems in national security contexts.
Balancing Innovation and Risk
While supporters argue that tools like Mythos could significantly strengthen national cybersecurity defenses, critics warn that exposing such powerful systems to government networks must be handled with extreme caution.
As discussions continue, the potential integration of advanced AI into federal cybersecurity operations signals a major shift in how governments may use artificial intelligence in the future.
