OpenAI's Cyber Defense Models Now Available on Amazon Bedrock

OpenAI has extended its cyber defense initiative, Daybreak, to Amazon Bedrock, making two specialized AI models available to eligible customers. These models, Daybreak Red and Daybreak Blue, are designed to assist cybersecurity teams in tasks such as vulnerability discovery, incident response, and exploit research.

Daybreak Red provides access to GPT-5.6 Cyber, a model trained specifically for advanced cybersecurity research, including vulnerability analysis and exploit reproduction. Daybreak Blue offers GPT-5.6 Sol, calibrated for defensive cybersecurity work like detection engineering and incident response.

Both models operate with zero-operator access enforced at the chip level, ensuring that customer code and sensitive data remain secure during inference. They run on Amazon Bedrock's next-generation inference engine, which guarantees high performance, security, and reliability. Customer data is encrypted both in transit and at rest, with access governed by AWS IAM policies and logged via AWS CloudTrail.

The models are already being used by AWS security teams to analyze source code and discover vulnerabilities. For example, security researchers utilized GPT-5.6 Cyber through Daybreak Red to identify previously unknown vulnerabilities in the JavaScript engine V8, which were subsequently fixed.

Availability is currently limited to the US East (N. Virginia) region, with access requiring enrollment in OpenAI's Trusted Access for Cyber program. Customers interested in deploying these models should contact OpenAI or their AWS account team for eligibility and setup.

This initiative underscores AWS and OpenAI's commitment to integrating security into AI infrastructure, providing cybersecurity professionals with powerful, secure tools to stay ahead of emerging threats.