← Evidence ledger
Capability gains

OpenAI releases GPT-5.6-Cyber through restricted Daybreak access

OpenAI released a cyber-specialized model that completes 95 percent of advanced dual-use requests, found high-severity real software vulnerabilities, and is available only to verified defenders under monitored Daybreak Red controls.

CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM69confidence 93/100

Why it moved the index

The model substantially reduces refusals for exploit development and has validated practical impact through previously unknown vulnerabilities in V8 and other real software. Identity verification, legal restrictions, monitoring, hardware-key requirements, and limited trusted access sharply constrain deployment, but the release still demonstrates a material expansion of advanced offensive-capable automation.

AUDIT TRAIL

Assessment history

  1. R1
    Toward 69 · confidence 93

    New historical specialized model release with real-world vulnerability findings and explicit safeguards.

    11 Aug 2026