Please do not create public GitHub issues for security vulnerabilities.

Report vulnerabilities privately via one of the following channels:

1. GitHub Security Advisory — navigate to the repository's "Security" tab and click "Report a vulnerability"
2. Contact the project maintainer directly — via the contact information available in the commit history or repository profile

• Repository name and branch
• Type of vulnerability (e.g., XSS, SQL injection, secret exposure, authentication bypass)
• Steps to reproduce (use synthetic data only — do not include real personal data in reports)
• Affected files or endpoints (if known)
• Impact assessment
• Suggested fix (optional)

Do not submit the following to any public repository or issue tracker:

If you discover any of the above in a public repository, report it via the vulnerability channels above — do not discuss it in public issues.

FNB is built on the principle that AI should not silently own, rewrite, or trade human memory and relationships.

Any change that allows AI-generated results to modify user-owned memory, relationship, permission, or identity objects must include:

1. Input reference — what input data was used
2. Model identity — which model produced the output
3. Confidence score — how confident the model is
4. AIInference record — structured inference ledger entry
5. Explanation record — human-readable explanation of the inference
6. User correction path — how the user can confirm, reject, or rewrite
7. Audit trail — full event history of the inference and its consequences

Plugins must declare:

• Data read scope — what data the plugin reads
• Data write scope — what objects the plugin creates or modifies
• External network behavior — whether the plugin contacts external services
• AI provider behavior — whether the plugin uses external AI models
• Storage behavior — whether the plugin stores data externally
• Audit behavior — whether the plugin writes audit records

Plugins must NOT:

• Exfiltrate user data without consent
• Train on user data by default
• Bypass permission boundaries
• Bypass audit records
• Bypass user correction (confirm/reject/rewrite)
• Write final Relationship objects without Evidence
• Write final Memory objects without MemorySource
• Externalize FNB Credit into tradable assets
• Use the FNB official brand identity without authorization

• All dependencies are scanned via Dependabot or GitHub dependency graph
• Pull requests that introduce new dependencies require maintainer review
• Unmaintained or abandoned dependencies should be replaced
• Lock files (go.sum, package-lock.json) must be kept in version control

This security policy covers:

• The FNB2026/fnb-open public showcase repository
• The FNB2026/fnb-protocol protocol and SDK repository
• Future selectively released engineering modules

It does not cover the private official implementation repository. Security issues related to private implementation code should be communicated through established internal channels.