← All pages

Input License

Definition: A clause covering what rights an AI provider claims over your Input — prompts, uploaded files, conversation content — ranging from narrow, purpose-limited use to maximal, perpetual, royalty-free, sublicensable grants. It’s the GenAI-dataset equivalent of the legacy platforms’ Content-license clauses (tracked in PGAv2), just applied to what you type into or upload to an AI product instead of what you post on a social platform. Risk level: Risk here varies by design — this keyword spans everything from this dataset’s most protective clauses (purpose-limited, training excluded) to its most aggressive (maximal, “for any purpose” grants stacked together). This wiki doesn’t score that range itself; see methodology.md §4 for why, and the website’s critic panel for wearetosed’s and ToS;DR’s independent scores where available.

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

Common Wording

By default, we do not train our models on your business data… unless you have explicitly opted in. — ChatGPT, Business Privacy Policy, 2026-01-08

Worldwide, nonexclusive, royalty-free license over Connector/App Responses, surviving termination; expanded in the 2025-12-17 rewrite to also grant rights to “prepare derivative works” and to extend the grant to the App’s own end users. — ChatGPT, Developer Terms, 2025-12-17

We may use your content to train our models… You can opt out of training through our privacy portal. — ChatGPT, Privacy Policy, 2026-01-21

Default opt-out training (“If you do not want us to use your Content to train our models, you have the option to opt out”) plus a separate worldwide right to use Content “to provide, maintain, develop, and improve our Services.” — 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

Default opt-out training on Inputs/Outputs, with safety-flagged and self-reported content always eligible regardless of opt-out. — Claude.ai, Privacy Policy, 2026-04-17

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

Chats, audio/Gemini Live recordings, screen-automation screenshots, and imported competitor-platform chats are all used “to improve our services (including training generative AI models),” gated by Keep Activity but with a carve-out that survives opt-out. — Google Generative AI Services, Privacy Policy, 2026-04-17

Perpetual, irrevocable, worldwide, royalty-free training license over Customer Data/Outputs, conditioned on tier and exception categories that have broadened with each revision. — Le Chat, Commercial Terms, 2026-04-08

“We may use Your Content to operate Copilot and improve it… we can copy, distribute, transmit, publicly display, publicly perform, edit, translate and reformat it.” — Microsoft Copilot, Terms of Service, 2026-01-13

“These rights and licenses are royalty free, transferable, sub-licensable, worldwide and irrevocable (for so long as Your Content is stored with us).” — Perplexity, Terms of Service, 2026-01-24

“Unconditional, irrevocable, non-exclusive, royalty-free, fully transferable, sub-licensable, perpetual, worldwide licence… to make derivative works of… any User Content,” plus an identical separate license over name/image/likeness. — Qwen Chat, Terms of Service, 2026-04-09

“An irrevocable, perpetual, transferable, sublicensable, royalty-free, and worldwide right to xAI to use, copy, store, modify, distribute, reproduce, publish, display in public forums… make derivative works of… your User Content… for any purpose.” — xAI, Terms of Service, 2025-12-11

Purpose-limited to providing the Services and safety/compliance/moderation operations, explicitly excluding training/product-improvement uses. — xAI, Commercial Terms, 2025-12-11

Notes & Trends

This keyword spans the widest risk range of any tag in the GenGA dataset. At the protective end, xAI’s Commercial Terms and ChatGPT’s Business Privacy Policy narrow the Input license to opt-in or purpose-excluded training. At the maximal end, Qwen Chat’s and xAI’s consumer Terms of Service grant perpetual, irrevocable, fully-stacked licenses “for any purpose” — rivaling the broadest PGAv2 social-platform grants (see genga_vs_pgav2_comparison.md). The clearest pattern: for every provider with both a consumer and an enterprise/API document (ChatGPT, Claude.ai, Le Chat, xAI), the enterprise tier’s Input license is narrower than the consumer tier’s — Anthropic’s and xAI’s Commercial Terms both make unconditional no-training commitments that neither provider makes in its consumer Terms of Service. This page was added retroactively after a wiki lint found 21 platform pages already using this keyword — well past the 3-platform cross-linking threshold in CLAUDE.md §4.3 — without a concept page for it; two pages (Le Chat’s Commercial Terms, Microsoft Copilot’s Terms of Service) already linked here in anticipation.