What happened
On Nov 4, 2026 OpenAI announced that Daybreak Red, access to the purpose-trained GPT-5.6 Cyber model, is now available to approved customers in AWS GovCloud (US-East) through Amazon Bedrock, for authorized vulnerability research, exploit validation and security testing. Inference data is not used to train models and is not shared with OpenAI; access requires OpenAI approval and AWS provisioning.
Why it matters
Expands frontier defensive cyber model availability to a government-grade, approval-gated cloud environment, giving public-sector defenders an additional lane for authorized vulnerability research and exploit validation.
Applicability
Government/defense security teams with AWS GovCloud and an approved Daybreak Red application should pursue enablement for authorized defensive testing.