BYOK Chat handles provider credentials and request content in transit, so security reports are treated as sensitive.

Security fixes target the latest commit on the default branch and the current deployment at https://byok.chat. Older commits, forks, and modified self-hosted deployments are not maintained by the BYOK Chat maintainers.

Do not open a public issue or pull request for a suspected vulnerability.

Use GitHub's private vulnerability reporting flow at:

https://github.com/heathermhuang/byok-chat/security/advisories/new

If GitHub private reporting is unavailable, email support@byok.chat with [BYOK Chat security] in the subject. Do not include live credentials or private prompt content in the first message.

Include:

• the affected route, component, or commit;
• the impact and realistic attack scenario;
• minimal reproduction steps;
• whether any real credential or personal data may have been exposed; and
• a suggested remediation, if you have one.

Use test credentials and synthetic content whenever possible. Never attach a live provider key, passphrase, prompt history, or another person's data to a report.

We aim to acknowledge reports within five business days. We will coordinate remediation and disclosure timing with the reporter based on severity and user impact.

• API-key exposure in browser storage, logs, analytics, exports, or error messages;
• cross-site scripting or script-policy bypasses that could read local workspace data;
• server-side request forgery or private-network access through custom endpoints or tools;
• provider credential forwarding to an unintended origin or redirect target;
• bypasses of attachment limits, tool permissions, rate limits, or consent gating; and
• vulnerable dependencies with a practical impact on the hosted application.

The public trust boundaries, deployment requirements, and non-goals are documented in docs/SECURITY_MODEL.md.

Self-hosters are responsible for their own Cloudflare account, domains, secrets, logging, private-network bindings, legal notices, and incident response.