A striking political reversal has unfolded concerning Anthropic and its advanced AI model, Mythos, following an earlier report on Project Glasswing. What was once deemed too dangerous for public release and positioned as a supply chain risk by the Trump administration, is now at the center of high-level discussions in the White House, driven by its unparalleled cybersecurity capabilities.

The shift in government stance was highlighted when Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent in the West Wing. While President Trump publicly claimed “no idea” about the visit, both the White House and Anthropic described the talks as “productive and constructive.” This meeting marks a significant change from just weeks prior when the Trump administration had designated Anthropic a “supply chain risk”—a label usually reserved for foreign adversaries—and President Trump himself stated the administration would “not do business with them again.”

A federal judge in San Francisco has since intervened, temporarily blocking the enforcement of that directive, thereby allowing Anthropic to continue working with non-military agencies while the legal battle continues. However, the dispute with the Pentagon remains active and unresolved.

The catalyst for this recalculation at the White House level is Anthropic Mythos AI's exceptional cybersecurity ability. Agencies are reportedly witnessing Mythos perform tasks no other tool can. The model, though not specifically trained for security work, developed an autonomous capacity to identify and exploit software vulnerabilities through general advancements in reasoning and code. During internal testing, Mythos uncovered thousands of previously unknown, high-severity vulnerabilities across major operating systems and web browsers, including a 27-year-old bug in OpenBSD and a 16-year-old flaw in FFmpeg that had eluded five million automated tests.

Rather than a public release, Anthropic made Mythos available to a select coalition of organizations—Project Glasswing—which includes industry giants like AWS, Apple, Cisco, Google, Microsoft, Nvidia, CrowdStrike, and JPMorganChase, supported by up to US$100 million in use credits. Within this controlled environment, Mythos is used