The Pulse
Anthropic Opens Mythos 5.1 to More Cybersecurity Teams
Anthropic is expanding its Cyber Verification Program into three tiers, giving qualifying security teams access to advanced Claude models, including Mythos 5.1. The new program sets different eligibility checks and safeguards for defensive
AI.info Team · Written with AI ·
Anthropic is widening access to its advanced cyber models while keeping the most sensitive uses behind deeper checks. On October 6, 2026, the company announced a three-tier Cyber Verification Program (CVP) that includes Claude Mythos 5.1 alongside Opus 5.5, Sonnet 5.5 and future models. Security teams can apply for the level that fits their work; the announcement does not open the models to general use.
The change brings together two existing programs. Project Glasswing gave a select group of organizations access to Mythos for work on critical software, while the earlier CVP gave vetted security teams access to Opus and Sonnet with reduced cyber safeguards. Anthropic says current Glasswing members will move to the new Specialized Access tier without reapproval for current models.
Three tiers, three levels of permission
Defense Access covers security operations, incident response, malware reverse-engineering, and vulnerability analysis and validation. Anthropic says eligible applicants can include security teams at companies, nonprofits, universities and government bodies, as well as critical-infrastructure operators, smaller security firms, open-source maintainers and individual researchers with a record of reported vulnerabilities. The company expects to respond to applications within a few days.
Red Team Access adds authorized penetration testing and red-teaming, but applicants must be organizations; individual researchers are not eligible. Tests must target systems the organization has permission to assess. Anthropic says it will still block actions that could cause physical harm or mass disruption, including ransomware deployment, damage to physical systems and penetration testing of high-risk safety systems. Reviews for this tier may take a few weeks, and applicants receive Defense Access while Anthropic evaluates their request.
Specialized Access has the fewest cyber blocks and is reserved for a limited set of verified organizations authorized to test systems whose failure could affect lives or disrupt markets. Anthropic names flight operating systems, power grids, telecom networks, interbank transfer infrastructure and government administrative networks as examples. The company says it reviews these organizations in depth with the U.S. government.
Anthropic’s safety test shows the trade-off
Anthropic tested the tiered controls on Claude Opus 5.5 using CyScenarioBench, an evaluation of multi-stage cyber operations under realistic constraints. The company ran five attempts on each of 10 challenges per tier. Without CVP access, every task was blocked on the first prompt; Defense Access blocked 46 of 50 trials at some point, while four succeeded.
In Red Team Access, no trials were blocked, and the model completed 34 of 50 tasks. Anthropic says that result is effectively equivalent to its 67.6% completion rate with no safeguards applied, which it uses as the comparison for Specialized Access. The test makes the program’s basic trade-off explicit: broader defensive work receives stronger intervention, while approved adversarial testing gets fewer blocks.
“The changes we’re making to our Cyber Verification Program today are intended to extend the impact of Project Glasswing to a much larger number of cyber defenders,” Anthropic writes in its announcement.
Glasswing’s reported findings underpin the expansion
Anthropic says Glasswing partners found at least 129,000 verified software vulnerabilities between April and July 2026. Its own open-source scanning efforts found another 5,500 verified vulnerabilities between April and October. More than 33,000 of the combined findings have so far been rated high- or critical-severity, the company says.
Anthropic cautions that the figures are a lower bound. The severity tally draws on partial data from 33 partner reports, and the company says it expects the actual impact to be at least five times higher. Fewer than half of partners disclosed how many vulnerabilities had been patched, often because fixes remained in progress.
Access comes with monitoring requirements
Organizations in CVP must allow data retention so Anthropic can monitor for cyber misuse. The company says organizations using Claude Fable 5.1 or Mythos 5.1 with zero data retention can also use CVP with that setting while its Enterprise Frontier Safeguards service is still in development. Anthropic says eligible organizations will be able to store data in cloud infrastructure they control once that service becomes available later this fall.
The program is available through the Claude Platform, Google Cloud Vertex AI and Microsoft Foundry. Amazon Bedrock access is limited to customers eligible for Enterprise Frontier Safeguards. Anthropic is accepting applications and says it will verify applicants and check that they have the security controls required for their requested tier.