OpenAI announces GPT-5.6-Cyber, a new AI model aimed at cybersecurity work, and it is made available through its Daybreak program. SecurityWeek and Quartz both report that access is provided to vetted defenders, while NDTV says it is delivered through Daybreak Red for trusted customers.
Quartz reports that in testing the model completes 95% of advanced cybersecurity task requests, indicating strong performance on difficult prompts. NDTV adds that OpenAI says the model handles advanced security tasks with fewer refusals and that internal tests show it outperforms general-purpose models. NDTV also names early access partners, including SpecterOps, SentinelOne and Palo Alto Networks, who reportedly use the model to improve defensive workflows.
Beyond the model itself, SecurityWeek says OpenAI expands the Daybreak platform to give more organizations access to its AI, aligning with Quartz’s description of restricted availability for vetted users. Across outlets, the emphasis is on defender-focused access, performance in cybersecurity testing, and broader roll-out of the Daybreak platform.
