Anthropic Scales AI Cybersecurity
Anthropic Widens Claude Mythos AI Cybersecurity Access to 150 New Global Partners
Anthropic is rapidly expanding its cybersecurity-focused AI initiative—rebranded as Claude Mythos under the Project Glasswing effort—by adding about 150 new organizations across more than 15 countries, bringing the partner pool to roughly 200. Access remains tightly controlled, with each organization required to meet Anthropic’s security requirements before it can use Claude Mythos Preview, and Anthropic says Mythos-class models will become broadly available to all customers in the coming weeks. The company credits collaboration with the U.S. government and members of the security industry for scaling the program and says early partners have reported finding more than 10,000 security flaws deemed highly or critically severe, while some experts argue fears of AI-enabled hacking have been exaggerated. The expanded partner set is intended to widen real-world testing across sectors including healthcare, power, water, communications, and hardware and includes additional high-profile organizations such as Samsung and NATO. The rollout follows Anthropic’s growing government and financial-industry attention and comes as it moves closer to an initial public offering filing, while external observers urge close attention to how disclosures, mitigation timelines, and testing coverage evolve with the larger network.
Key Points
Anthropic said it extended access after “several weeks of close collaboration with our Project Glasswing partners, the security industry, open-source software maintainers, and the US government,” and reiterated that each new organization “will need to meet our security requirements before they gain access.”
Following Anthropic’s April Mythos Preview announcement, U.S. officials reportedly escalated the issue to banks: Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell “convened an urgent meeting with bank CEOs to warn of the risks,” as financial institutions raced to patch vulnerabilities.
Anthropic declined to name the new organizations it is adding, but said the expansion includes government organizations and estimated that a major attack on partner organizations could impact “more than 100 million people.”
In its initial rollout, Anthropic shared Claude Mythos access with only select partners (including Apple), first expanding to about 50 organizations in April specifically to test the model’s ability to detect cybersecurity vulnerabilities that leave enterprise systems exposed.
Sentiment
Mostly PositivePublishers
23
Articles
70
Reach
93
