OpenAI reports a remote code execution flaw in Brotli, a compression library used in Node.js and Envoy
OpenAI says it found an unauthenticated remote code execution vulnerability in Brotli during a red teaming exercise. According to the company, new versions and a CVE are already available, but each piece of software that bundles the library must be updated separately.

Photo: Thiện Ân · Dominio público · via Flickr
OpenAI said on its official X account that, during a red teaming exercise, it found an unauthenticated remote code execution (RCE) vulnerability in Brotli, a widely used compression library. The post is dated October 23. The information below comes from that post, not from independent verification.
What OpenAI says
According to the company, Brotli is included in software such as Node.js, Bun and Envoy. The flaw would allow an attacker to run their own code on affected systems by sending, or returning as a response, malicious compressed data. No user account would be required.
OpenAI describes two scenarios:
- Unpatched server that accepts compressed data: a malicious client could take control.
- Unpatched client that accepts compressed data: a malicious server could take control. The example cited is `curl --compressed`.
Why the fix is not immediate
OpenAI explains that the root cause had already been fixed in Brotli's source code in February. At the time, it was not known to create a security vulnerability, and no subsequent release of the library followed.
Because Brotli is embedded in other programs, each of them must incorporate the update. Users of those programs must then install the new versions.
Disclosure status
OpenAI says it notified Brotli and projects that depend on it, such as Node.js. According to the company, those projects have already released or are releasing security updates. Brotli has reportedly released new versions and published a CVE. The post points to patchtheplanet.org/brotli for details on patches and mitigations.
What to do
The material does not specify version numbers or the CVE identifier. For that reason, the prudent course for technical teams is to:
- Check whether their applications, servers or proxies use Node.js, Bun, Envoy or other software that includes Brotli.
- Consult each project's security advisories and the mitigation page linked by OpenAI.
- Consider both services that receive compressed data and clients that request it.
What remains unclear
The material does not detail what the red teaming exercise involved or how the flaw was discovered. Nor does it say whether there is active exploitation, what severity has been assigned, or which exact versions are affected or fixed. The cited post appears to be part of a broader thread about the security of OpenAI's training environment, but the available text is cut off.
Source: OpenAI on X