OpenAI and AWS announced the next step in their partnership on Tuesday: OpenAI's cybersecurity models, known as Daybreak, are now available through Amazon Bedrock — inside the AWS environments where enterprises already build, secure and operate software.
Through the 'Daybreak Access' program, approved customers get two levels: Daybreak Blue provides frontier general-purpose models such as GPT-5.6 Sol, with safeguards tailored to authorized defensive security work; Daybreak Red offers purpose-trained cybersecurity models for authorized vulnerability research, exploit validation and security testing.
The models are designed to accelerate the work security teams do under pressure — from vulnerability discovery and detection engineering to incident response, including complex workflows like exploit reproduction and mitigation development. Access is available through the Bedrock console or the Responses API.
The announcement lands in a turbulent week for OpenAI. The company recently acknowledged that it cannot rule out a 'critical' cybersecurity capability level in its upcoming Astra model and paused parts of its development. At the same time, OpenAI and its rivals are pushing into AI-powered defense — a market observers see as the commercial counterpart to the offensive capabilities of modern models.
For AWS, the deal is another step toward embedding AI startups into the enterprise cloud; for OpenAI, it is a channel to sell cybersecurity AI through the governance, procurement and security processes of large companies. Both companies stress that access must remain controlled and authorized — giving security teams the tools, without handing the same power to attackers.
