ai-disclosure-standard/SECURITY.md

Security policy

Version 1.6.x is the supported line in this project archive.

Reporting

Report suspected vulnerabilities privately to the project operator before public disclosure. Configure the operator's security contact outside this source archive and publish it through the project website or a security.txt file at the deployment edge.

Secure defaults

• HTML responses use a restrictive Content Security Policy with a fresh nonce.
• Request IDs are length- and character-validated before they enter logs.
• Query strings and client IP addresses are not written to application logs by default.
• Forwarded client addresses are trusted only from explicitly configured proxy CIDRs.
• /metrics is disabled by default and requires a Bearer token when enabled.
• JSON validation requests are size-limited, strictly decoded and concurrency-limited.
• The reference container runs without root, Linux capabilities or a writable root filesystem.

See docs/LEGAL-AND-SECURITY.md for deployment controls that remain the operator's responsibility, including TLS, edge rate limits, patching, secret management, network segmentation and log deletion.

License integration

• Customer installations configure only LICENSE_TOKEN and optional client settings.
• LICENSE_PUBLIC_KEY, private keys and signing keys are not supported.
• Public issuer and lease keys are embedded from internal/app/trusted_keys.json at build time.
• Key generation, license issuance, token registries, revocation and lease signing exist only in the standalone Universal License Platform.
• Use HTTPS for hybrid and online verification.
• Protect the hybrid cache directory from other local users; it contains signed lease tokens, not private keys.

Online mode fails closed if the platform is unavailable. Hybrid mode may continue only while a previously signed lease remains valid within the grace period encoded in the license.