OpenAI says GPT-5.6 Cyber is now available to approved customers in AWS GovCloud (US-East) through Daybreak Red. The company announced the expansion on November 4, 2026. It builds on the earlier availability of Daybreak models on Amazon Bedrock, and it targets government security teams that work inside AWS environments.
This is an access story, not a new model. The model, GPT-5.6 Cyber, is already part of Daybreak Red. What changed is where approved customers can reach it. Here is what the announcement says, what the earlier AWS launch added, and what the approval process looks like.
What OpenAI announced
In its announcement, OpenAI says the expansion "gives government security teams a new way to apply advanced cybersecurity capabilities within the AWS environments they use to support critical missions."
The post describes Daybreak Red as access to "OpenAI's purpose-trained cybersecurity models for authorized vulnerability research, exploit validation, and security testing." It says GPT-5.6 Cyber can help defenders investigate complex software weaknesses, reproduce vulnerabilities in controlled settings and develop mitigations.
OpenAI adds a caution: "Qualified security practitioners should review findings and proposed fixes before implementation." The model is a research aid, and the post puts human review before any change ships.
Why government teams need a separate path
OpenAI argues that for government organizations, adopting advanced cybersecurity capabilities "requires more than model performance." Teams also need to work within established security, governance and operational requirements.
Availability in AWS GovCloud (US-East) gives eligible teams a path to bring GPT-5.6 Cyber into those workflows through Amazon Bedrock. OpenAI says defenders can use it to reason through security findings, assess their impact and evaluate fixes, which helps connect vulnerability research to practical improvements in defense.
The announcement does not name any government customer, agency or deployment. It also does not give pricing, usage limits or performance figures for the GovCloud offering. It is an availability notice. The AWS model card for the model does list AWS GovCloud (US-East) as an in-region option with its own Standard-tier prices, so check that page for current rates.
What "fewer refusals" means
General-purpose AI models often decline requests that look like attack work, such as reproducing a vulnerability. Daybreak Red takes a different approach. OpenAI says it "gives verified defenders greater latitude for authorized cybersecurity work, with fewer refusals for these tasks alongside stronger identity verification, monitoring, and access controls."
The AWS announcement from August explains the reasoning in its own words. AWS says cybersecurity is "inherently dual use," and that a request to reproduce a vulnerability or reverse engineer an exploit chain "looks identical regardless of intent." AWS says Daybreak Red and Daybreak Blue resolve that ambiguity "through context: who is using the model, where the work occurs, and what safeguards govern that access."
In short, the stated trade is fewer refusals for stronger identity verification, monitoring and access controls.
Red and Blue
AWS describes two products on Bedrock. Daybreak Red provides access to GPT-5.6 Cyber, which AWS calls a purpose-trained cybersecurity model. Daybreak Blue provides access to GPT-5.6 Sol with safeguards calibrated for defensive cybersecurity work.
According to AWS, Daybreak Blue is "the right starting point for most security teams." It supports vulnerability discovery, detection engineering and incident response. Daybreak Red is designed for advanced tasks like vulnerability research, exploit reproduction and mitigation development.
OpenAI's GovCloud announcement covers Daybreak Red only. It does not say whether Daybreak Blue is available in GovCloud.
Both models are part of Daybreak, which AWS describes as the cyber defense initiative from OpenAI that gives defenders governed access to frontier AI, including agentic tooling, application red teaming, and services that help defenders move from findings to tested fixes.
Data handling
For Daybreak Red on Amazon Bedrock, OpenAI says "inference data is not used to train models, and customers do not need to share their Daybreak Red data with OpenAI."
The AWS post, written for the Bedrock launch, adds more detail on that setup. It says:
- Zero-operator access is enforced at the chip, so even AWS operators cannot access prompts and completions during inference.
- Data is encrypted in transit and at rest.
- Access is governed by AWS Identity and Access Management policies, logged in AWS CloudTrail and routed through virtual private cloud endpoints.
- For automated abuse detection, classifier-flagged traffic is retained by AWS for up to 30 days and processed programmatically. Customers may request zero data retention through their AWS account team.
Those details come from the August Bedrock post, which lists US East (Ohio) as the region. OpenAI's GovCloud post does not repeat them for GovCloud, so a GovCloud customer should confirm the terms with its AWS account team.
How access works
Access is gated in two steps. OpenAI says access "requires OpenAI Daybreak Red approval and AWS provisioning." To start, complete the Daybreak application or contact your OpenAI or AWS account team about eligibility. Once approved, you work with your AWS account team to enable access.
One line is easy to miss: "Daybreak Red approval does not automatically enable model access on AWS." Approval from OpenAI and provisioning from AWS are separate steps, and a team needs both.
The AWS model card for the model says access is limited to eligible customers. It lists the supported use cases as vulnerability research, exploit reproduction and mitigation development. The August AWS post adds that access requires enrollment in Trusted Access for Cyber from OpenAI.
What was shown earlier
In its August post, AWS says that, according to OpenAI, security researchers used GPT-5.6 Cyber through Daybreak Red to identify two previously unknown vulnerabilities in V8, the JavaScript engine used by Chrome. AWS says the two, when chained together, could enable memory corruption and a heap sandbox escape. It says the initial vulnerability was fixed and released as CVE-2026-15903.
That claim is secondhand. It comes from AWS, citing OpenAI, and the GovCloud announcement does not repeat it. Treat it as an example of the kind of work the product targets, not as a measure of how it will perform in any particular agency.
OpenAI has also described large-scale vulnerability work in other settings. For one example, see our post on OpenAI's pipeline that found 948 Linux kernel privilege escalation flaws. For another security-focused AI move, see Cognition Is Acquiring Corridor for Devin Security.
What to watch
The announcement leaves several questions open. It does not say how many agencies or contractors are approved. It does not say whether the GovCloud offering has different limits from the commercial region. It also does not describe compliance authorizations, so teams with specific requirements will need to ask AWS and OpenAI directly.
For now, the news is narrow and clear. GPT-5.6 Cyber, first listed on Bedrock in the US East (Ohio) region, is now also available to approved customers in AWS GovCloud (US-East). Anyone who wants to use it still needs both OpenAI approval and AWS provisioning, and OpenAI says qualified practitioners should review findings and fixes before implementation.
Sources
- Daybreak Red is now available in AWS GovCloud (US-East), OpenAI, November 4, 2026
- Accelerate cyber defense with OpenAI and AWS: Daybreak Red & Daybreak Blue now available to eligible customers on Amazon Bedrock, AWS Machine Learning Blog
- Daybreak Red: GPT-5.6 Cyber, Amazon Bedrock model card
- Cognition Is Acquiring Corridor for Devin Security, llms.blog
- OpenAI Finds 948 Linux Kernel Privilege Escalation Flaws, llms.blog
