OpenAI reorganizes ChatGPT and Codex

OpenAI on May 11 launched Daybreak, a cybersecurity platform that combines GPT-5.5 and Codex tools for defensive security work, according to the company’s product page. The release puts the company’s chatbot, coding agent and API-era model lineup into a more unified offering for one high-risk use case: finding vulnerabilities, generating patches and verifying fixes in code and live systems. Anthropic has been building a parallel push around Claude Mythos Preview and Project Glasswing, with both companies emphasizing vetting, monitoring and restricted access rather than broad public release. ### What exactly did OpenAI put under Daybreak? OpenAI’s Daybreak page says the product is designed to “identify threats, generate patches, and verify remediation” across code and systems. The page describes a package built around GPT-5.5 and “Codex Security,” with workflow steps for triage, patch generation and audit-ready verification. GPT-5.5 became available in ChatGPT, Codex and the API in late April, according to OpenAI’s product announcement. (openai.com) In that release, OpenAI said the model was built for complex tasks including coding, research and tool use, and said nearly 200 early-access partners had provided feedback before launch. Codex itself has moved from a standalone coding agent toward a broader engineering product. OpenAI’s current Codex page describes it as an agentic coding system that can plan, build, review and release software across tools, while earlier product updates said GPT-5-Codex was available through the Responses API and ChatGPT subscriptions. (openai.com) ### Where does the “reorganization” show up? April 23 is the clearest date in OpenAI’s public materials tying the product lines together. (openai.com) OpenAI said then that GPT-5.5 was rolling out to Plus, Pro, Business and Enterprise users in ChatGPT and Codex, while GPT-5.5 Pro was also being rolled out in ChatGPT and the API. September 2025 and April 2026 updates show the same convergence from the developer side. OpenAI said GPT-5-Codex was available through API keys in the Responses API, and later added flexible team pricing for Codex in ChatGPT Business and Enterprise. (openai.com) Those moves placed the coding agent, subscription product and developer platform on overlapping rails rather than separate tracks. That framing is an inference from OpenAI’s product changes, not language the company itself used. (openai.com) ### How is Anthropic positioning Mythos? Anthropic last month introduced Claude Mythos Preview through its red-team research site and Project Glasswing initiative. The company said Mythos represented a “substantial leap” in cyber capabilities and said the model had already found thousands of high-severity vulnerabilities, including some in major operating systems and web browsers. Project Glasswing lists Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks as launch partners. (openai.com) Anthropic also said it was committing up to $100 million in usage credits for Mythos Preview and $4 million in direct donations to open-source security organizations. ### Why are both companies talking so much about access controls? (red.anthropic.com) OpenAI last week described “Trusted Access for Cyber” as an identity- and trust-based framework for verified defenders. The company said approved users receive fewer classifier-based refusals for authorized workflows such as vulnerability triage, malware analysis, reverse engineering, detection engineering and patch validation, while harmful requests remain restricted. (anthropic.com) Anthropic has used similar language around controlled deployment. Its Glasswing materials say the company has been discussing Mythos Preview and its cyber capabilities with U.S. government officials, and frame the program as an effort to place advanced capabilities with defenders responsible for critical software. ### What can users actually do next? OpenAI’s next step is explicit on its Daybreak and Trusted Access materials: vetted defenders and teams can apply for access through the company’s cyber program. (openai.com) Anthropic’s next step is also public: organizations can review Project Glasswing and Claude Mythos Preview materials, including the named partner list and research write-ups on Anthropic’s sites. (openai.com) (anthropic.com)