AI Copyright Indemnity Checker — Will Your AI Provider Defend You?

This is a lookup-and-decision tool, not a calculator. Each AI provider is a record with one or more plan tiers, and every tier stores facts taken from that provider's published legal terms: whether it offers a copyright indemnity (the promise to defend you and pay the costs of a third-party IP claim), what that indemnity covers, which content types it applies to, whether the payout is capped, and the conditionsyou must meet. The tool never guesses: where a provider is silent, the tier reads "Not stated" and asks you to check the terms.

Your content type and intended use then decide the verdict chip. The derivation is deterministic and applied in order:

1. If the tier offers no indemnity → Not covered. This is every consumer/free tier here — ChatGPT Free/Plus, consumer Copilot, the Gemini app, Claude Free/Pro, GitHub Copilot Individual, Midjourney, Stability AI and self-hosted open-weights models.
2. If the terms are silent → Not stated(for example third-party models running on Amazon Bedrock, which follow their own providers' terms rather than the AWS indemnity).
3. If your content type is outside what the indemnity covers → Not covered, and the card names the types that are covered. GitHub Copilot's commitment, for instance, covers generated code, not images.
4. If the indemnity is commercial-only and you picked personal use → Not covered. These shields exist to back business customers, not hobby projects.
5. If the indemnity is conditional → Conditionally covered, and the conditions are listed — most commonly "keep the content filters and guardrails on" and "don't try to generate infringing content."
6. Otherwise → Covered. Even here the standing conditions are shown, because every real indemnity requires guardrails-on and no willful infringement.

To keep the dataset honest, the build runs two internal cross-checks. The first re-derives every tier's verdict a second way — mapping the indemnity flag to a base verdict, then applying the content-type and commercial-only gates through an independent code path — and asserts the two agree across all 19tiers, four content types and three uses. The second checks internal invariants, such as "a tier with no indemnity must not claim a payout cap." The "10 providers · 19tiers cross-checked" badge means both pass. Everything runs in your browser over a static, dated snapshot — no API calls, no scraping of live terms.

• OpenAI — Business Terms & Copyright Shield (API + ChatGPT Enterprise/Team)
• Microsoft — Customer Copyright Commitment (Copilot, Azure OpenAI)
• GitHub — Copilot Product-Specific Terms (Business/Enterprise IP indemnity)
• Google — Generative AI indemnification (Cloud / Vertex AI, two-part)
• Adobe — Firefly legal FAQ (enterprise IP indemnification)
• Amazon Web Services — Service Terms (uncapped IP indemnity for Amazon-built model output)
• Anthropic — Commercial Terms (indemnification for authorized use of Services/Outputs)
• Midjourney — Terms of Service (liability on the user, no indemnity)
• Stability AI — Community & Membership License (no indemnity)

Each provider's indemnity fields were last cross-checked against its governing terms on 2026-07-17. Provider terms change often — this page tracks the current published version, not a historical changelog, and reflects standard published terms rather than negotiated enterprise agreements. It does not cover patent or trademark disputes, or country-specific case law, and is not a substitute for legal advice.