OpenAI launched GPT-5.6-Cyber on August 10 as part of a major expansion of its Daybreak cybersecurity initiative. The new model, available exclusively through the Daybreak Red tier, answered 95% of advanced security requests in internal testing—compared to just 1.5% for the standard GPT-5.6 Sol model with safeguards enabled.
Daybreak now offers two access levels: Blue, which provides GPT-5.6 Sol with system-level cyber guardrails removed for authorized defensive work, and Red, which unlocks GPT-5.6-Cyber for exploit validation, zero-day research, and red teaming. Starting September 1, all individual Daybreak accounts will require physical security keys, with access gated by identity verification, monitoring, and legal attestations.
Real vulnerabilities, vetted access only
OpenAI used GPT-5.6-Cyber to discover two previously unknown Chrome V8 vulnerabilities that could be chained to escape the V8 heap sandbox, disclosed to Google as CVE-2026-15903. The company also reported finding hundreds of other vulnerabilities across operating systems and databases. Partners like IBM, CrowdStrike, and Palo Alto Networks can now integrate Daybreak models into security products, but the underlying models stay with the approved partner—not the customer.
The move comes days after OpenAI paused its upcoming Astra model for triggering critical-level cyber risk in safety testing. Both GPT-5.6 Sol and Cyber fall under the high tier, below Astra's critical rating.