OpenAI's Daybreak Red and Blue land in Amazon Bedrock

OpenAI has brought its Daybreak cybersecurity models into Amazon Bedrock, making both access levels, Daybreak Blue and Daybreak Red, available inside customers' existing AWS environments. The rollout was announced by OpenAI as the next step in its work with AWS.

At a glance

• Access runs through Daybreak Access enrollment: once approved, eligible customers call the models from the Amazon Bedrock console or through the Responses API at the bedrock-mantle endpoint.
• Daybreak Blue covers frontier general-purpose models, including GPT-5.6 Sol, with safeguards scoped to authorized defensive work, while Daybreak Red covers models purpose-trained for vulnerability research, exploit validation and security testing.
• OpenAI positions the models across vulnerability research, detection engineering and incident response, from initial discovery through a validated fix, and for workflows such as exploit reproduction and mitigation development.

Why it matters Cyber capability is among the more tightly gated parts of OpenAI's lineup, and routing it through Bedrock moves that gate into a procurement and identity stack enterprises already operate. For most buyers that likely counts for more than raw capability, since a model that cannot clear a security review never reaches production. The arrangement also reads as a test of whether cyber-capable models can be distributed through a cloud platform under tiered access.

OpenAI says adopting specialized cybersecurity capabilities takes more than model performance, naming security review, governance, procurement, access controls and an operating model that teams can support. Delivery through Amazon Bedrock, by its account, puts the models inside the AWS security, governance and operational workflows enterprises already use.

The move follows the general availability of OpenAI frontier models and Codex on AWS earlier this year, which gave enterprises a route to run those systems in production. Daybreak is the next piece to arrive through that channel, this time addressed to defenders.

Both levels sit behind enrollment in Daybreak Access and are open to eligible customers. Daybreak Blue provides frontier general-purpose models, including GPT-5.6 Sol, with safeguards tailored to authorized defensive security work. Daybreak Red provides purpose-trained cybersecurity models for authorized vulnerability research, exploit validation and security testing.

OpenAI frames the span of work from initial discovery through a validated fix, listing accelerated vulnerability research, detection engineering and incident response. Heavier tasks are also named, among them exploit reproduction and mitigation development, which OpenAI describes as complex workflows the models support.

What's next Enrollment in Daybreak Access remains the gate for both levels, and approved customers reach the models from the Amazon Bedrock console or through the Responses API on the bedrock-mantle endpoint, with setup covered in OpenAI's documentation. The announcement names no pricing, no regional coverage and no eligibility criteria, and it does not say whether further Daybreak capabilities will follow on AWS.