The Race — Building Democracy in the Age of AI. People race toward a shared finish line.
The Race
← Trends

Supporting case

OpenAI gives a vetted group access to GPT-6 Cyber

What happened

OpenAI released GPT-6 Cyber to a subset of vetted Daybreak Red customers on November 5. It rated the model Critical in cybersecurity, the same capability category as GPT-6 Astra. Organizations need identity verification and controls including single sign-on, multifactor authentication and approved network addresses for API use. The model can help authorized defenders find and test vulnerabilities. Company and partner tests in controlled environments do not establish how the safeguards perform across real deployments. System card ↗ · Release ↗.

The earlier Astra and Daybreak records already establish differentiated access. This release shows that a new capable model is being distributed through the same provider-administered gate. It does not yet show that independent oversight, access rights or selection remedies changed.

Dates and implementation stage

Later action 2026-11-05; publication 2026-11-05

Supporting sources

Related trends