So basically flagging a notable change for anyone tracking this: Anthropic updated their Terms of Service on March 12, 2026. The key revision is in Section 4 (Ownership of Content). The new language explicitly states that “outputs generated by Claude models are not claimed as intellectual property of Anthropic, and the User retains all rights permitted by applicable law.” The phrase “permitted by applicable law” is new and important — it acknowledges the copyright registrability gap that @smalltownlegal_3 has been discussing throughout this thread.
Additionally, Section 7 now includes a new “Enterprise Indemnification” clause for API customers on the Scale or Enterprise tiers. Anthropic will indemnify enterprise customers against third-party IP infringement claims arising from Claude’s outputs, up to the amount paid under the contract in the prior 12 months. This is huge — it mirrors what OpenAI introduced in late 2024 and what Google offered for Gemini. For businesses building products on Claude, this significantly reduces legal risk lol.
One practical implication: if you are a freelancer or agency using Claude commercially through the consumer plan (Pro), you do NOT get indemnification. That protection is API-only and tier-restricted. The consumer ToS still say Anthropic makes “no warranty regarding the accuracy, completeness, or legal protectability of outputs.” So the risk allocation remains different depending on how you access the model. If your business depends on Claude outputs, moving to the API with an enterprise agreement is increasingly the correct legal posture.