Anthropic Expands Access to Mythos Cybersecurity AI Model Anthropic’s latest move to widen access to its powerful cybersecurity AI, Claude Mythos, is accelerating a global shift in how critical infrastructure is defended—while also stoking anxiety over dependence on a single private model.

In early April, Anthropic launched Project Glasswing with roughly 50 initial partners, who used Claude Mythos Preview “to scan their codebases for vulnerabilities,” uncovering “more than 10,000 high- or critical-severity security flaws.” The model is designed to autonomously identify and even exploit software weaknesses, with Anthropic pitching the project as a way “to secure the world’s most important software.”

By June 1, EU pressure over unequal access had come to a head. After weeks of negotiations, Anthropic agreed to grant the bloc’s cybersecurity agency ENISA access to Mythos, making it “the first EU institution to join Project Glasswing” and ending “one of the most visible flashpoints in the transatlantic AI relationship.” Earlier Financial Times reporting had framed this as part of broader talks in which “Anthropic offers EU access to Mythos” in its first expansion outside the US and UK.

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 sectors such as “power, water, healthcare, communications, and hardware.” TechCrunch summarised the move as scaling Claude Mythos “to critical infrastructure in 15+ countries,” targeting systems where a cyberattack “could affect 100 million people.” The Verge likewise noted that “Anthropic is giving Claude Mythos Preview to around 150 more organizations,” particularly in industries “that weren’t well represented” previously.

From Anthropic’s perspective, the expansion is a controlled safety measure: each partner “will need to meet our security requirements before they gain access,” and the goal is to help cybersecurity adapt before “cheap, fast AI models with powerful cyber capabilities are around the corner.” But some industry voices are uneasy about the company’s growing centrality. In a widely shared X post, AI pioneer Yann LeCun amplified criticism that Anthropic “thinks it’s building God” and called that ambition “the ultimate level of narcissism and delusion of grandeur.”

As rivals such as OpenAI launch competing initiatives and governments race to secure critical systems, Anthropic’s Mythos stands at the intersection of urgent security needs and deep concern over concentrated AI power.

1. “Expanding Project Glasswing” – Anthropic blog describing Project Glasswing’s goals, initial results, and the plan to extend access to 150 organizations in over 15 countries.
2. “Anthropic gives EU cybersecurity agency ENISA access to Mythos AI” – Report detailing ENISA’s access, Mythos’s discovery of 10,000+ zero-day vulnerabilities, and the end of a weeks-long EU–US standoff.
3. “Anthropic offers EU access to Mythos” – Financial Times coverage of talks to make Mythos available to EU institutions beyond the US and UK markets.
4. “Anthropic scales Claude Mythos to critical infrastructure in 15+ countries” – TechCrunch article on expanding Project Glasswing to 150 organizations across critical infrastructure sectors.
5. “Anthropic is giving Claude Mythos Preview to around 150 more organizations.” – The Verge report highlighting new access for underrepresented industries like power, water, and healthcare.
6. @ylecun on X – Yann LeCun retweeting criticism that Anthropic “thinks it’s building God,” calling it “the ultimate level of narcissism and delusion of grandeur.”

Continue reading https://foxvector.com/stories/019e9068-4e99-31ed-70ef-2107db5dd01e