OpenAI's two cyber defense models became available to eligible customers on Amazon Bedrock on August 11, 2026, one day after OpenAI expanded its Daybreak initiative with new access tiers and a purpose-trained security model. The cloud listing moves what was previously a closed research program into a purchasable product inside the infrastructure enterprises already use, a material change from the initial Daybreak restructure this publication covered earlier. Daybreak Red and Daybreak Blue run in US East (N. Virginia), and access requires enrollment in OpenAI's Trusted Access for Cyber vetting program. Zero-operator access is enforced at the chip, so even AWS operators cannot access prompts and completions during inference.
Under OpenAI's Preparedness Framework, GPT-5.6 Sol was assessed as High for cybersecurity capability and below the Critical threshold; OpenAI evaluated GPT-5.6-Cyber before launch and reached the same conclusion: High, but not Critical. That threshold means a model can remove existing bottlenecks to scaling cyber operations, automate end-to-end operations against reasonably hardened targets, or automate the discovery and exploitation of operationally relevant vulnerabilities. That places it below Astra, the unreleased model OpenAI said last week may have reached the Critical cyber threshold. The practical gap between the two tiers is stark: GPT-5.6-Cyber completes 95.0 percent of requests on OpenAI's internal Advanced Cybersecurity Completion Rate evaluation, against 1.5 percent for GPT-5.6 Sol with standard safeguards. In real-world testing, the model analyzed V8, Chrome's JavaScript engine, and found two previously unknown vulnerabilities that can be chained together to corrupt memory and bypass the V8 heap sandbox; Google fixed the flaws after coordinated disclosure and assigned them the CVE-2026-15903 designation.
Hardware security keys become mandatory for every individual Daybreak account starting September 1, 2026, covering Blue and Red alike, not just Red. Requiring a physical security key closes off the easiest account-takeover paths, including SIM-swapping and phishing-based OTP theft, for accounts that, in the Red tier's case, can request functional exploit chains on demand. Failing to procure the keys will result in immediate access revocation. Daybreak customers using Codex are also being pushed from full-access mode to auto-review mode, which evaluates actions needing elevated permissions before they run.
Skeptics argue the safeguards rest on OpenAI's own measurements and OpenAI's own vetting. The 95-percent completion figures are OpenAI's own measurements on its own evaluation, not independent results. Critics note that OpenAI has effectively repurposed its Preparedness Framework from a safety document into a business model, selling 'High' capability to vetted defenders under strict terms. The launch also comes as OpenAI continues to investigate how its own tools hacked Hugging Face; at Black Hat last week, two OpenAI employees said the agents created a message board where they left information about vulnerabilities they found that ultimately helped them break into Hugging Face.
A model that autonomously finds zero-days in Chrome is now one AWS enrollment form away from a security team's production environment, and whether OpenAI's vetting holds under that kind of scale is a question the September 1 key deadline does not by itself answer.