# Security and responsible disclosure

Please report vulnerabilities or unsafe model behavior through a private
message to the repository owner before opening a public discussion with
reproduction details.

Do not include access tokens, private audio, personal data, or unredacted
speaker recordings in reports. Revoke any credential that is accidentally
published.

This project supports voice-conditioned synthesis. Applications must obtain
the reference speaker's informed consent, disclose synthetic audio where
appropriate, and prevent impersonation, fraud, harassment, and
voice-authentication bypass.