Skip to content

Security: eai-support/help

SECURITY.md

Security Policy

Policy owner role: Security
Status: approval-pending
Last reviewed: 2026-07-31
Next review due: 2026-10-31
Review cadence: quarterly and after a material change

Supported scope and versions

The supported product scope and versions are approval-pending. Do not infer support from repository visibility, release history, an open Issue, or this draft policy. Customer promotion remains blocked until Security approves and publishes the exact scope.

Report a vulnerability privately

Use EnterpriseAI's approved GitHub private vulnerability reporting route. The launch renderer supplies the reviewed route and its isolated security-owned fallback before this policy is published.

Do not open or post vulnerability details in a public Issue or Discussion. Do not include exploit material, credentials, customer or tenant information, confidential diagnostics, or security-sensitive attachments in a public channel. Public editing, transfer, locking, or deletion does not restore confidentiality after disclosure.

If private vulnerability reporting is unavailable, use only the approved restricted-private fallback displayed by the published security route. Do not substitute a public Issue, Discussion, personal address, or unreviewed endpoint.

What to expect

The human acknowledgement target is approval-pending. This policy does not promise a response, remediation, disclosure, or release time until Security and Legal approve explicit wording. Report state and technical detail remain inside the restricted security workflow.

Security may coordinate a customer-safe advisory after validation and approval. Public advisories never expose reporter identity, confidential evidence, or private remediation work.

Accidental public disclosure

Stop ordinary triage and invoke the sensitive-disclosure runbook. Preserve only the minimum reference needed for response, rotate exposed credentials when relevant, request GitHub sensitive-data removal, and publish only an approved sanitized explanation.

There aren't any published security advisories