OpenAI Expands Daybreak Cybersecurity Program With New Tiers

OpenAI announced on Monday, August 10, 2026, a significant expansion of its Daybreak cybersecurity research program, introducing two new access tiers and a specialized AI model designed to help defenders identify and test software vulnerabilities before attackers can exploit them. The move aims to provide security professionals with advanced tools to counter the growing threat of AI-powered cyberattacks.

The expansion includes the launch of Daybreak Blue and Daybreak Red, two distinct access levels within the program. Daybreak Blue is positioned as the starting point for most defenders, allowing software teams to scan code for vulnerabilities, validate patches, and analyze malware. It operates with reduced safeguards compared to OpenAI's standard commercial models. For more advanced red-teaming and penetration testing, OpenAI introduced Daybreak Red, which provides access to a new model variant called GPT5.6Cyber.

GPT5.6Cyber is a specialized version of OpenAI's flagship GPT5.6 Sol model, specifically optimized for cybersecurity tasks such as discovering zero-day vulnerabilities and building exploit chains. In an internal benchmark measuring the completion of advanced...