OpenAI has brought its Daybreak Red service to AWS GovCloud, giving approved government security teams access to GPT-5.6 Cyber through Amazon Bedrock. OpenAI published its access guidance on 4 November 2026, following AWS's own regional availability announcement on 30 October. The service is aimed at authorised cybersecurity work, not general use, and access is tightly gated.

Two approvals, not one

Using the service takes two separate approvals. An organisation must first be approved through OpenAI's Trusted Access for Cyber programme, and AWS must then enable the model on the customer's account. OpenAI states plainly that Daybreak Red approval does not automatically enable model access on AWS. The offering covers AWS GovCloud (US-East) only, with in-region inference and no cross-region routing, and it is available only to approved customers in that region.

What GPT-5.6 Cyber is for

GPT-5.6 Cyber is a purpose-trained cybersecurity model based on GPT-5.6 Sol, built for vulnerability research, exploit validation and security testing. It carries a 1,050,000-token context window and is identified in Bedrock as openai.gpt-5.6-cyber. Reporting on the announcement notes that both OpenAI and AWS say inference data for Daybreak Red on Bedrock is not used to train models, and that flagged traffic may be retained for up to 30 days for abuse detection unless zero data retention is requested.

Why GovCloud matters

AWS GovCloud is an isolated cloud region built for US government agencies and other regulated workloads. Daybreak models first reached Amazon Bedrock in US East (Ohio) in August 2026; the GovCloud expansion moves the same governed-access model into the environment where public-sector security teams already operate. The public announcements name no government customer, and they do not claim the service is open to every AWS account.