Anthropic has folded Project Glasswing into an expanded Cyber Verification Program with three access tiers, giving vetted security teams its most capable models, including Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1, with fewer cyber blocks at each step up1,2. The company announced the change on October 6 and said new models will be added to the tiers as they ship.
Because Anthropic treats cybersecurity as dual use, its generally available models carry conservative safeguards that block most cyber work. When Opus 5.5 launched on September 22, most security tasks sent to it were routed to the older Opus 4.8. Until now, trusted access ran through two programs: Glasswing, which since April had given organizations securing critical software access to Mythos, and the original verification program, which gave vetted teams reduced safeguards on Opus and Sonnet models.
ANALYSIS The change turns Mythos access from a partner program for a group of critical-software organizations into one that any qualifying security team can apply for, with blocks set by tier.
Three tiers
Defense Access covers defensive work such as incident response, malware reverse engineering and vulnerability analysis. It is open to security teams at companies, nonprofits, universities and government bodies defending their own systems, critical-infrastructure operators as small as a regional hospital, smaller security firms, open-source maintainers and individual researchers with a record of reported vulnerabilities; Anthropic aims to respond to applications within a few days.
Red Team Access adds authorized penetration testing and red-teaming, limited to systems an organization is authorized to test and, for now, to organizations rather than individual researchers. Real-time blocks still apply to actions that could cause physical harm or mass disruption, such as deploying ransomware; reviews take a few weeks, and applicants work in the Defense tier in the meantime.
Specialized Access, which has the fewest cyber blocks, is reserved for a limited set of verified organizations authorized to test safety systems such as flight operating systems, power grids, telecom networks, interbank transfer infrastructure and government administrative networks. Anthropic reviews each organization in depth with the U.S. government, and existing Glasswing members move into the tier without reapproval for current models. In a statement reported by Bloomberg, the company said those with access can carry out "high risk offensive testing" of safety systems that protect critical infrastructure such as power grids, banks and flight operating systems3.
What the tests show
Anthropic tested the tiers on CyScenarioBench, an evaluation of multi-stage cyber operations, running Opus 5.5 five times on each of its 10 challenges in every tier. Without program access, every attempt was blocked on the first prompt; in the Defense tier, 46 of 50 runs were blocked at some point4. In the Red Team tier nothing was blocked, and the model completed 34 of 50 runs, which the company said is effectively equivalent to its 67.6% success rate with no safeguards applied.
Glasswing's record
Anthropic said Glasswing partners uncovered at least 129,000 verified vulnerabilities between April and July, and that its own open-source scanning found 5,500 more between April and October; more than 33,000 of them have been rated critical or high severity. It called the figures a lower bound drawn from 33 partner reports, estimated the true impact at least five times higher, and said fewer than half of partners disclosed how many flaws they had patched.
VulnCheck researcher Patrick Garrity was less impressed: of the 225 Anthropic-linked vulnerabilities he tracked, fewer than 0.5 percent were being exploited in the wild, The Register reported. The publication also noted that the company's own figures show 516 of 5,674 true-positive vulnerabilities patched.
Organizations in the program must allow data retention so Anthropic can monitor for misuse; Enterprise Frontier Safeguards, due later this fall, will let eligible customers keep that data in cloud infrastructure they control. The program is available on the Claude Platform, Google Cloud's Vertex AI and Microsoft Foundry, and on Amazon Bedrock only to customers eligible for Enterprise Frontier Safeguards.