Daybreak Cyber Defense Models Land on Amazon Bedrock

OpenAI's Daybreak cyber defense models have landed on Amazon Bedrock. The two models, Daybreak Red and Daybreak Blue, are now available to eligible customers. According to AWS, these models will run under the same infrastructure controls as every other critical workload.

Daybreak Red provides access to GPT-5.6 Cyber, a model trained specifically for cybersecurity tasks like finding zero-day vulnerabilities. In contrast, Daybreak Blue offers GPT-5.6 Sol with safeguards recalibrated for defensive security work such as vulnerability discovery and detection engineering.

AWS Vice President John Sheehan stated that their teams are using both models to analyze source code, discover vulnerabilities, and conduct red-team research on Bedrock.

OpenAI's Daybreak expansion post revealed numbers on how much the refusal behavior shifts across the tiers. GPT-5.6 Cyber through Daybreak Red completes 95.0% of requests involving exploit-chain development, authentication bypass, and privilege escalation, compared to 2.0% for GPT-5.6 Sol under Daybreak Blue access.

OpenAI says its researchers used GPT-5.6 Cyber to investigate V8, the JavaScript engine inside Chrome, and uncovered two previously unknown vulnerabilities that could be chained to corrupt memory and escape the V8 heap sandbox.