AI Usage Policy

This Policy governs AI features across the Services. Auric orchestrates interactions; it does not train, host, or own the models. Capitalized terms follow Auric's Master Definitions in the Terms of Service.

• Explicit invocation: AI runs only when you trigger it. Nothing is sent in the background.
• Your own tools, your own provider: Auric holds project state and does not sit between you and a model. There is no AI provider SDK in the product and no call site — nothing in Auric sends a prompt anywhere.
• What is sent: only the content you include in a request, which may contain portions of Project Data you choose to share.
• What is not sent: Auric does not upload your project data or operational state. Server-side, workspaces hold a repository name and remote URL — never file contents.

Once content reaches a provider, that provider's terms, privacy policy, retention, and any model-training practices apply. Review them. Auric cannot retrieve or delete content held by a provider.

Generated Output may be inaccurate, incomplete, biased, or unsafe. You must review and validate it before use — especially for code that is executed, deployed, or relied upon. Auric makes no representation about AI correctness or fitness.

Ownership and permitted use of Generated Output are determined by the applicable provider's terms and by law. Auric claims no ownership of your prompts or outputs.

You must not use connected AI to produce unlawful, infringing, harmful, or provider-prohibited content, or to evade provider or Auric controls. See the Acceptable Use Policy.

If you work against a local model, nothing leaves your device on that path either; you remain responsible for that model's licensing and behaviour.

If these terms change materially we will update the effective date on this page and, where the change affects you meaningfully, tell you by email before it takes effect. Contact: hello@auric.cx.

2 September 2026: The beta is described as private throughout, which is how access is actually granted — requested, confirmed, then opened in waves. No right, obligation, party or data practice changed.

1.3.0-beta.1 (2026-08-02): Published for the private beta. Review markers resolved against the implementation, unsupported AI-provider claims removed, and retention stated from the constants that enforce it.