Security policy
Cahut is a server that takes HTML from an AI and serves it to readers. Its security matters more than any feature. Thank you for reporting what you find.
Reporting a vulnerability
Report privately, never in a public issue or discussion:
- through GitHub’s private vulnerability reporting: report a vulnerability;
- or by e-mail to
security@cahut.cloud.
Say what you found, how to reproduce it, and what you think it allows. English or French.
What to expect
- An acknowledgement within seven days.
- A fix and a coordinated publication within ninety days, much sooner for a serious flaw. You know the date before we publish.
- A GitHub security advisory with a CVE, and a patch release that cites it.
- Credit in the advisory and in the release notes, unless you prefer not to be named.
There is no bug bounty.
Scope
In scope: the cahut binary and its commands, the MCP endpoint, the console and its API, install.sh, the release pipeline, and the sites cahut.cloud and docs.cahut.cloud. Any way for a Window to read another Window, the console or the reader’s session is a vulnerability: the sandbox is the main protection (ADR 10).
Out of scope: the content that people publish in their Windows, the AI clients, and third-party services.
Supported versions
Only the latest release receives fixes. Cahut is 0.x: update as soon as a release is out. A server updates itself every day to the patch releases of its minor version, and sudo /usr/local/bin/cahut update installs any newer release.
Reports written by an AI
Reproduce before you report. A report generated by an AI that nobody has verified is closed without an answer. A fabricated report gets its author banned.