We take the security of AiDotNet seriously. If you believe you have found a security vulnerability, please report it via one of the channels below.

Report privately through the GitHub Security Advisory form. This is the fastest path and gives us a structured audit trail.

If you cannot use GitHub Security Advisories, email admin@aidotnet.dev.

For sensitive disclosures, GitHub Security Advisories above is the encrypted path. Direct email currently has transport-layer confidentiality only — a dedicated AiDotNet PGP key will be published at https://aidotnet.dev/security once generated, and this section will be updated with the published key fingerprint at that time.

Please do not report security vulnerabilities through public GitHub issues, discussions, or any other public forum.

We aim to meet the following response times:

Severity uses the CVSS 3.1 calculator. We assign CVE identifiers via GitHub Security Advisories for any vulnerability rated SEV-2 or higher.

We follow a 90-day embargo from the date we acknowledge the report, or until a fix is publicly released — whichever is sooner. Reporters who prefer a shorter or longer embargo for coordinated disclosure should say so in their initial report. We will not unilaterally extend embargoes beyond 90 days without reporter agreement.

We provide security fixes for the following versions:

Until AiDotNet reaches 1.0, supported versions are the two most recent minor releases. After 1.0, we will publish a long-term-support window and update this table accordingly.

Security policy applies to:

• AiDotNet (main library)
• AiDotNet.Tensors (numerical core)
• AiDotNet.Serving (inference server)
• AiDotNet.Native.* (native acceleration packages)
• Pre-built NuGet packages distributed via nuget.org under the AiDotNet org

Out of scope (report directly to the upstream maintainer):

• Vulnerabilities in third-party dependencies — please report to the upstream project first; we will coordinate downstream remediation once upstream has a fix or assigns a CVE.
• Vulnerabilities in Pre-trained Community Models that we distribute as metadata only — report to the model author per their model card.

We credit reporters in our advisory text and in release notes unless the reporter requests anonymity. If you wish to publish your own write-up after the embargo ends, please share the draft with us at least 48 hours in advance so we can coordinate timing.

For U.S. government, military, national laboratory, and other regulated deployments (FedRAMP, FISMA, NIST AI RMF / SP 800-218 SSDF, CMMC, HIPAA, SOC 2 Type II, ISO 27001, etc.), please contact admin@aidotnet.dev for our Enterprise security program. The Enterprise tier includes:

• Air-gapped deployment support (no telemetry, no license-server callout)
• FIPS 140-3 compatible cryptographic modules
• SBOM (CycloneDX 1.5) generation per release
• SLSA Level 3 build provenance attestations
• Signed NuGet packages
• Dedicated security contact + custom SLA
• NIST SP 800-218 SSDF compliance documentation

See https://aidotnet.dev/federal-use for the federal-use page or https://aidotnet.dev/enterprise for general enterprise terms.

This project is currently maintained by a single primary maintainer. If you have not received an acknowledgement within 5 business days via the primary channels above, please escalate to admin@aidotnet.dev with subject [ESCALATION] Security report not acknowledged. We are actively working to add a backup security contact and will update this section once that is in place.

We will not comment publicly on a vulnerability before the embargo expires or a fix is released. Once the embargo lifts, our public statement will appear on the GitHub Security Advisories page for the report.