← All pages

Output Ownership

Definition: A clause stating who owns the AI-generated Output a user receives — almost always framed as user-protective, with the provider disclaiming ownership and/or assigning any rights it holds to the customer, in contrast to the broader, less protective Input License clauses often found in the same documents. Risk level: Low, and generally protective — this is one of the few keywords in the GenGA dataset that consistently favors the user rather than flagging a risk. It’s tracked here for completeness and cross-provider comparison, and because this protective framing is often paired with — and can obscure — a much broader Input license in the same document.

Note on provenance: This concept page covers the sources/GenGA/ (GenGA/Open Terms Archive) dataset, which has no pre-tagged JSONL — all tagging here is LLM-assigned by direct reading. See each linked platform page’s methodology note.


Platforms Using This Clause

PlatformDocument TypeDateLink
ChatGPTCommercial Terms2025-12-18link
ChatGPTTerms of Service2026-01-16link
Claude.aiCommercial Terms2026-03-11link
Claude.aiTerms of Service2025-11-11link
DeepSeekDeveloper Terms2026-03-27link
DeepSeekTerms of Service2026-03-27link
Google Generative AI ServicesTerms of Service2026-04-15link
Le ChatCommercial Terms2026-04-08link
Le ChatTerms of Service2026-04-16link
Llama APITerms of Service2026-03-20link

Common Wording

Customer: (a) retains all ownership rights in Input; and (b) owns all Output. OpenAI hereby assigns to Customer all OpenAI’s right, title, and interest, if any, in and to Output. — ChatGPT, Commercial Terms, 2025-12-18

You (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output. — ChatGPT, Terms of Service, 2026-01-16

Customer (a) retains all rights to its Inputs, and (b) owns its Outputs… Anthropic may not train models on Customer Content from Services. — Claude.ai, Commercial Terms, 2026-03-11

Input ownership retained by developer; Output rights assigned to developer; both subject to a broad explicitly-permitted use list including third-party model training/distillation. — DeepSeek, Developer Terms, 2026-03-27

Identical Input/Output ownership-and-permissive-use clause as the Developer Terms, explicitly permitting “training other models (such as model distillation).” — DeepSeek, Terms of Service, 2026-03-27

Google won’t claim ownership over user-generated content. — Google Generative AI Services, Terms of Service, 2026-04-15

User retains Input ownership and owns all Output; Mistral assigns any Output rights it might hold to the user. — Le Chat, Terms of Service, 2026-04-16


Output ownership is the most consistently user-protective clause type in the GenGA dataset: every provider tracked here disclaims ownership of, or affirmatively assigns, Output rights to the user or customer. This consistency is itself worth flagging as a pattern, not just a list of reassurances: nearly every page in this table pairs its protective Output-ownership clause with a same-document Input License clause that is considerably broader. DeepSeek’s and Le Chat’s Output-ownership disclaimers, for example, sit directly alongside permissive third-party training-distillation rights and, for Le Chat, a license that eroded from an unconditional no-training baseline into a multi-exception opt-out structure. Read in isolation, an Output-ownership clause can give a misleadingly reassuring impression of a document’s overall risk — a same-document Input-license or Output-protection clause is frequently the more consequential one. (This point was previously framed in terms of risk-scoring weights; risk scoring was retired 2026-06-21 — see methodology.md — so it now stands as a reading caution rather than a scoring rule.) This concept page was added retroactively after a wiki lint identified 10 platform pages using this keyword without a concept page existing, despite being well past the 3-platform threshold in CLAUDE.md §4.3.