Anthropic Expands Mythos AI Cybersecurity Model to 150 New Partners Across 15 Countries

CNBC reported Tuesday that Anthropic is opening its Mythos cybersecurity model to an additional 150 organisations across more than 15 countries. The rollout, conducted under the company’s Project Glasswing initiative, follows a wave of inbound requests from sectors that were absent from the program’s debut. New entrants span power, water, healthcare, communications, and hardware industries.

Mythos Cybersecurity Model Gets New Rules for Entry

Joining partners will not receive automatic access. Anthropic confirmed that each organisation must satisfy defined security criteria before the Mythos cybersecurity model is handed over. The company framed the requirement as essential given the model’s sensitivity. In a blog post cited by CNBC, Anthropic described the expansion as a step toward a broader ambition: making all software inherently more secure while helping the industry rethink cybersecurity assumptions in an AI-driven era.

Also Read: OpenAI Launches GPT-5 With Extended Reasoning and Multimodal Upgrades

How Project Glasswing Began

Anthropic first tested Mythos with a limited cohort of 50 partners in April, after internal reviews surfaced concerns about the model’s potency. The worry was straightforward: a tool powerful enough to find software vulnerabilities could equally empower malicious actors. Those concerns triggered discussions among the White House, major technology companies, and large financial institutions before any wider release was approved. The cautious initial rollout was a direct response to that pressure.

Also Read: What the AI Arms Race Means for Enterprise Security Spending

Partners Have Already Found Thousands of Flaws

Since the April launch, the existing cohort of Project Glasswing partners has surfaced more than 10,000 high or critical security vulnerabilities, according to Anthropic. That figure lends weight to the company’s argument that controlled, wide distribution of the model produces measurable defensive gains. The results appear to have satisfied internal thresholds enough to justify a much larger rollout.

Expansion Arrives at a Pivotal Moment for Anthropic

The Mythos news lands during an especially active stretch for the startup. One day before Tuesday’s announcement, Anthropic disclosed it had confidentially filed its IPO prospectus with the Securities and Exchange Commission, moving ahead of rival OpenAI on that milestone. The company also separately confirmed it would extend Mythos access to partners inside the European Union. Together, the moves signal an aggressive push to establish Anthropic’s security-focused AI as a global standard before competitors can consolidate their own positions in the enterprise market.

Read Next: Anthropic Files Confidential IPO Prospectus With the SEC