Anthropic expands Mythos cybersecurity model access to 150 partners
Anthropic expanded access to its Mythos cybersecurity AI model from 50 to 150 partners across 15+ countries to identify software vulnerabilities, including sectors such as energy and healthcare. Experts had warned that the same capabilities could potentially be misused by attackers to find security flaws more quickly.
What happened
Anthropic announced that 150 additional partners across more than 15 countries would gain access to Mythos, an advanced AI model for detecting software vulnerabilities, under Project Glasswing.
The expansion includes sectors previously underrepresented in the initial phase, such as energy, water, health, telecommunications, and hardware. Initial participants started with 50 partners in April 2026.
Reported impact
According to Anthropic, Project Glasswing participants have identified over 10,000 high or critical severity vulnerabilities since the program's start.
Anthropic estimated that a large-scale cyberattack could affect more than 100 million people for most of the newly participating organizations.
Index analysis
Experts had warned that the same capabilities enabling Mythos to identify vulnerabilities could be exploited by malicious users to find software security gaps faster.
Anthropic stated that within the next year, more AI companies will possess equally powerful models and could release them without safety guardrails to prevent misuse. The company wrote that cyberattacks could then occur much more frequently and in less predictable forms.
Governance controls
New Project Glasswing participants must meet specific security requirements before gaining access to the model.
The White House conducted a series of meetings with leading technology executives and financial institutions to examine ways to safely utilize advanced AI models.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.