Anthropic folds Project Glasswing into an expanded three-tier Cyber Verification Program
What happened
Anthropic expanded its Cyber Verification Program into a three-tier system that gives vetted security teams stepped-up access to its AI models. Each tier reduces the restrictions Anthropic places on cybersecurity research using its products. This change folds in Project Glasswing, a prior Anthropic initiative, to better accommodate ethical hackers and security researchers working with Anthropic’s AI.
Why it matters
Anthropic’s AI models carry strong safety and security classifiers because the company treats cybersecurity research as dual-use—it wants to prevent malicious use but still encourage legitimate security testing. This three-tier program balances those tensions by allowing trusted researchers to move beyond the usual product limits. For operators and cybersecurity teams, it clears the path for deeper vetting and testing that was previously blocked or limited by Anthropic’s safeguards. This matters because practical security work often needs more access and fewer usage hurdles to identify vulnerabilities or exploit vectors effectively.
What to watch next
Look for how Anthropic’s three-tier approach influences other AI providers managing the dual-use challenge between safety and security research. Will competitors adopt similar layered programs to let vetted teams do more extensive testing? Also watch how this shift affects the pace and quality of AI risk discovery. More researcher access could accelerate finding weaknesses but may pressure Anthropic to tighten oversight or update its classifiers. The program’s rollout and community response will reveal whether this new Cyber Verification Program strikes the right operational balance.
AI Quick Briefs Editorial Desk