• Early April: Mythos quietly rolls out
• May–June: Pressure builds for wider, international access
• June 2: Glasswing triples in size
• Competing visions and criticism

Anthropic Expands Access to Mythos Cybersecurity AI Model Anthropic is rapidly widening access to its powerful cybersecurity AI, Mythos, betting that sharing offensive capabilities with select partners will ultimately make global infrastructure safer even as critics warn of concentration of power and techno‑hubris.

§Early April: Mythos quietly rolls out

In early April, Anthropic launched Project Glasswing, giving “roughly 50 initial partners” access to Claude Mythos Preview to scan their codebases for vulnerabilities. Those partners used the model to uncover “more than 10,000 high- or critical-severity security flaws,” Anthropic said, positioning Mythos as a benchmark tool for discovering zero‑day vulnerabilities.

§May–June: Pressure builds for wider, international access

As results emerged, European officials grew concerned that Mythos was finding bugs in software used by their banks and infrastructure while no EU body could see the details. After “a weeks-long standoff,” Anthropic agreed to give the EU’s cybersecurity agency ENISA access, making it “the first EU institution to join Project Glasswing.” The company describes Mythos as able to “autonomously identify security flaws in complex codebases” and simulate attacks that once required teams of human researchers.

In parallel, the EU entered talks for broader adoption of the American model, in what would be Mythos’s “first expansion outside US and UK” markets. The move intensified debates in Europe about reliance on foreign AI infrastructure for critical security functions.

§June 2: Glasswing triples in size

On June 2, Anthropic formally announced it was “expanding Project Glasswing” to “approximately 150 new organizations” in “more than 15 countries,” with a focus on power, water, healthcare, communications, and hardware providers whose codebases, if breached, could each affect “more than 100 million people.” TechCrunch reported that Anthropic is “scal[ing] Claude Mythos to critical infrastructure in 15+ countries,” including U.S. allies such as Australia, Canada, France, Germany, India, Japan, and South Korea.

The Verge likewise noted that Anthropic is giving “Claude Mythos Preview to around 150 more organizations,” specifically to industries “that weren’t well represented” in the initial cohort, so they can use the model “to find security vulnerabilities.”

§Competing visions and criticism

Industry observers frame the expansion both as a security initiative and a strategic business play, as Anthropic’s fundraising and valuation accelerate alongside rivals OpenAI and Google. At the same time, public critics question the company’s ambitions. In a widely shared post amplifying investor Bill Gurley, Yann LeCun relayed the claim that “Anthropic Thinks It’s Building God,” calling it “the ultimate level of narcissism and delusion of grandeur to think you can create God.”

Anthropic, for its part, says the next phase of Glasswing is about preparing for an era in which “cheap, fast AI models with powerful cyber capabilities are around the corner,” and ensuring those tools strengthen, rather than destabilize, global cybersecurity.

Continue reading https://foxvector.com/stories/019e8a9d-c102-2555-72f2-2faea0e83cc9