Anthropic Expands Access to Claude Mythos Cybersecurity AI Anthropic is rapidly scaling its Claude Mythos cybersecurity AI from a tightly controlled pilot into a global tool for defending critical infrastructure, even as questions grow about dependence on a single, powerful private model.

In early April, Anthropic launched Project Glasswing, giving “roughly 50 initial partners” access to Claude Mythos Preview to scan their codebases for vulnerabilities. The company says those partners have already found “more than 10,000 high- or critical-severity security flaws.” Mythos, described as Anthropic’s “most powerful yet” cyber-focused model, can autonomously discover thousands of zero-days over several weeks.

Weeks of testing with major tech firms and U.S. government entities set the stage for a contentious geopolitical turn. European officials pressed for access after learning Mythos had uncovered vulnerabilities in software used by EU banks, governments, and infrastructure, while European institutions could not see the results. Over the weekend before June 1, Anthropic agreed to give the EU’s cybersecurity agency ENISA access, making it “the first EU institution to join Project Glasswing” and ending “a weeks-long standoff” in transatlantic AI relations.

On June 2, Anthropic formally announced it was “expanding Project Glasswing” to “approximately 150 new organizations” across “more than 15 countries,” with a focus on power, water, healthcare, communications, and hardware providers whose compromise “could be catastrophic” and affect “more than 100 million people.” TechCrunch reported that the expanded group targets “critical infrastructure in power, water, healthcare, and communications where a cyberattack could affect 100 million people,” and includes entities such as NATO, Okta, Samsung, and ENISA.

Consumer tech and business media framed the move as a major scale-up: “Anthropic is giving Claude Mythos Preview to around 150 more organizations,” The Verge noted, while the Financial Times highlighted plans to “expand Mythos access to more than 15 countries” and offer the EU broader use of the model.

Even as Mythos becomes the benchmark for AI-driven vulnerability discovery, with rival initiatives like OpenAI’s Daybreak emerging, some industry voices worry about the ambition behind such systems. In a widely shared clip, investor Bill Gurley argued that “Anthropic Thinks It’s Building God,” calling it “the ultimate level of narcissism and delusion of grandeur to think you can create God,” a message amplified by AI pioneer Yann LeCun on X.

Anthropic, for its part, pitches Glasswing as a defense-first project to “secure the world’s most important software” and prepare cybersecurity for an era of cheap, fast AI models with powerful offensive capabilities.

Continue reading https://foxvector.com/stories/019e91b2-08c1-1d12-716b-2f11a9ef5797