OpenAI Expands Frontier Cyber Model Access to Daybreak Partners

OpenAI launched the Daybreak Cyber Partner Program on June 22 to let security vendors and consultancies embed its frontier models in products, managed services and customer engagements. The initiative addresses a widening defense gap as attackers use AI to find vulnerabilities and develop exploits faster, while many organizations lack the expertise or infrastructure to operate specialized cyber models themselves. Daybreak pairs advanced model access with established security teams, governance and human oversight.

OpenAI said on Aug. 10 that approved partners can use Daybreak Blue for broad defensive work or the more tightly governed Daybreak Red for red teaming and penetration testing. Partners named include Accenture, IBM, EY, Palo Alto Networks, CrowdStrike and Cloudflare; underlying model access stays with the partner rather than the customer. GPT-5.6-Cyber completed 95.0% of advanced cybersecurity requests in an internal test, versus 1.5% for GPT-5.6 Sol and 2.0% under Blue access. Financial terms were not disclosed.