OpenAI’s Daybreak and Anthropic’s Glasswing have nearly identical benchmarks — and 3 of the same partners

OpenAI launched Daybreak, a tiered cybersecurity program built on GPT-5.5 variants for defensive and authorized offensive security work.

Anthropic had already introduced Project Glasswing, a vetted consortium using Claude Mythos Preview for similar security tasks.

The two initiatives are closely aligned in goals and workflow, highlighting how leading AI labs are converging on cybersecurity offerings.

Cisco, CrowdStrike, and Palo Alto Networks are involved in both programs, suggesting major vendors are testing both ecosystems rather than choosing one.