Anthropic has reorganized its external security initiatives, folding Project Glasswing and the Cyber Verification Program into a new three-tier threat hunting club. The reconfiguration, reported by The Register, marks a shift from separate programs to a single, tiered structure for security researchers.
Details of the individual tiers have not been disclosed, but the move suggests Anthropic is looking to standardize how it works with outside threat hunters. The company appears to be consolidating its outreach efforts under one umbrella rather than maintaining distinct programs with separate goals.
For the compute and hardware angle, the change is notable because AI security increasingly depends on coordinated external testing of models and infrastructure. A tiered club could give Anthropic a clearer pipeline for prioritizing vulnerabilities and rewarding researchers, though the source does not specify any technical or hardware-related details.