# Cloudflare uses OpenAI Daybreak to investigate vulnerabilities and propose fixes, and Codex to speed up development

> Cloudflare is using OpenAI's Daybreak to investigate vulnerabilities and propose fixes for review, while its developers use Codex to build faster and improve code quality. The public account is thin on detail.
>
> OpenCheese AI · 2026.10.26 · Web page: https://opencheese.dev/blog/cloudflare-uses-openai-daybreak-to-investigate-vulnerabilities-and-propose-fixes

Finding a vulnerability is the easy part of security. Everything after that (working out what it means, deciding what to change, getting the change reviewed) is where the calendar fills up. Cloudflare says it now uses OpenAI tools to shorten that second half of the story.

## What the report actually says

The source for this story is a short summary from OpenAI, and it makes two claims.

- **Daybreak for security work.** Cloudflare uses OpenAI Daybreak to investigate vulnerabilities and propose fixes for review.
- **Codex for everyday development.** Cloudflare's developers use Codex to build faster and improve code quality.

That is the whole of it. The headline frames the first point as linking vulnerability discovery to defense, which is a tidy way of saying that the gap between "we found a problem" and "here is a candidate fix" is something the company wants to close.

## The part worth noticing: fixes go to review

The wording matters. Daybreak, as described, proposes fixes. It does not, according to the summary, ship them. The phrase "for review" places a checkpoint between the tool's suggestion and the production system, which is where most security teams would want one.

The summary does not say who does the reviewing, how long it takes, or how often a proposed fix is accepted unchanged. Those are the numbers that would tell you whether this is a real shift in how a security team works or a nicer way to draft patches. None of them are in the material available, so none appear here.

## Two tools, two jobs

The report draws a quiet line between two kinds of work. Daybreak sits on the defensive side: investigate, then propose. Codex sits on the building side, where developers use it to move faster and, in the report's words, improve code quality.

It is a sensible pairing on paper. Code gets written, vulnerabilities get found in code, and fixes become more code. A company that writes a great deal of software has an obvious interest in tools at both ends of that loop. Whether the two are connected inside Cloudflare, or simply used by different teams, the summary does not say.

## What we do not know

The summary is short, and it is also a company's own account of a customer relationship. A few things are absent:

- Any measured results, such as time saved or vulnerabilities resolved.
- How Daybreak is deployed, or which of Cloudflare's systems it touches.
- How Codex is used day to day, or what "improved code quality" is measured against.
- Any statement from Cloudflare itself.

Until those details appear, the fair reading is modest. Cloudflare is a named user of both products, and the stated use cases are plausible. The evidence for how well they work is, for now, an adjective.

## The takeaway

The interesting claim is not that an AI tool can look at vulnerabilities. It is that its output goes to a person for review before anything changes. That is the sort of boring safeguard that makes a security tool usable, and it is also the sort of detail that usually gets cut from a two-sentence summary. A fuller account would be welcome.

## Related posts

- 2026.10.28 · [CodeRabbit puts GPT-6.1 Sol to work hunting security risks in large codebases](https://opencheese.dev/blog/coderabbit-gpt-6-1-sol-security-agent-complex-codebases.md): OpenAI says CodeRabbit uses GPT-6.1 Sol to investigate security risks across complex codebases. Its security agent helps developers review the findings and the proposed fixes.
- 2026.10.28 · [Greptile extends bug detection past code review with GPT-6 Astra](https://opencheese.dev/blog/greptile-extends-bug-detection-past-code-review-with-gpt-6-astra.md): Greptile says it uses GPT-6 Astra to test user workflows, Daybreak to strengthen security and Codex to reduce rework. The published summary has no results or numbers.
- 2026.10.06 · [OpenAI Reports Second AI-Agent Access Incident in an Australian Government System](https://opencheese.dev/blog/openai-reports-second-ai-agent-access-incident-australian-government.md): An OpenAI agent accessed a New South Wales parks service application and retrieved non-public fire statistics. The company says its review found no personal information was obtained, while state authorities investigate.