Security

Built to be trusted in production.

Open source, self-hostable, and reviewed by teams running AI in regulated environments. Security is a default, not a feature.

Principles

How we think about security.

Data ownership

Self-host the framework. AINative never sees your prompts, responses, or API keys. There is no SaaS layer to bypass.

Secret handling

Provider API keys stay server-side. The client never receives credentials. Tools execute in your environment, with your permissions.

Transport security

TLS everywhere. SSE and WebSocket transports support cert pinning, custom headers, and signed requests.

Self-hosting

Runs in your VPC, on-prem, or air-gapped. No outbound calls required. No telemetry, ever.

Auditability

Open source. Every line of the runtime is inspectable on GitHub. Pinned, semver-versioned releases.

Compliance

Materials available for SOC 2, ISO 27001, and HIPAA reviews. Provided to Enterprise customers.

Practices

What we do, in detail.

Vulnerability disclosure

Report security issues to security@ainative.dev. We respond within 24 hours and publish CVEs through GitHub Security Advisories.

Dependency policy

Zero runtime dependencies in the client core. Server adapters use only well-maintained, audited packages with automated updates.

Release signing

All npm and PyPI packages are signed and provenance-attested. Verify with sigstore or your registry's UI.

Reproducible builds

CI builds are deterministic. Lockfiles are committed. You can reproduce any release locally.

Need security documentation?

Enterprise customers receive a complete security pack including SOC 2 review materials, pen-test summaries, and a vendor questionnaire.