Anthropic Expands Access to Mythos Cybersecurity AI Model Anthropic’s once tightly controlled cybersecurity AI, Mythos, is rapidly moving from a US- and UK-focused pilot to a globally shared tool at the center of transatlantic debates over digital sovereignty and AI risk.

In early April 2026, Anthropic quietly launched Claude Mythos Preview inside “Project Glasswing,” giving roughly 50 initial partners — including parts of the U.S. government — access to scan critical codebases for vulnerabilities. Those partners went on to uncover more than 10,000 high‑ or critical‑severity security flaws in widely used software. Mythos, described as Anthropic’s “most powerful yet” cybersecurity model, can autonomously identify zero‑days and generate working exploits at high success rates, compressing months of expert work into hours.

As news of these capabilities spread, frustration grew in Europe. ENISA, the EU’s cybersecurity agency, and euro‑area institutions learned that Mythos had found vulnerabilities in systems underpinning European banks, governments, and critical infrastructure — but no EU body could see the results. After weeks of contentious talks that became a flashpoint in EU‑US AI relations, Anthropic agreed to grant ENISA access, making it the first EU institution inside Project Glasswing and ending the standoff.

On June 1, reporting emerged that “Anthropic offers EU access to Mythos,” framing the ENISA deal as the model’s first expansion beyond the US and UK and a test case for broader EU adoption.

By June 2, Anthropic moved from selective concessions to a full scale‑up. In a company update titled “Expanding Project Glasswing,” it announced it would extend access to “approximately 150 new organizations” in “more than 15 countries,” focusing on power, water, healthcare, communications, and hardware providers where a successful cyberattack “could be catastrophic” and affect “more than 100 million people.” TechCrunch reported that the new cohort includes NATO, ENISA, Okta, and major Asian and European technology firms across countries such as Australia, Germany, India, Japan, and South Korea.

Human analysts highlight two competing narratives: for Anthropic, Project Glasswing is a “collaborative effort to secure the world’s most important software” and prepare cybersecurity for an era of cheap, powerful AI models; for European policymakers, the ENISA deal underscores lingering worries about relying on a US‑based AI system as the benchmark in defensive cyber operations. Both sides, however, now accept that Mythos and similar tools will shape how nations defend — and potentially attack — the digital infrastructure that underpins modern life.

Continue reading https://foxvector.com/stories/019e9d49-3ff7-31c4-7397-0c114b1bf942