1. What the document is: Anthropic's consumer-facing Terms of Service for Claude.ai.
What this wiki found — complete, every page
Presence facts, not verdicts — each flagged term links to its definition and the exact clause on this page. Absence of a badge means the term isn't currently flagged here, not that the page is risk-free.
Claude.ai – Terms of Service
Dataset: GenGA (Generative AI Governance Archive) — 11 AI services, 2025–present
GenGA (Generative AI Governance Archive) is this wiki’s Generative-AI dataset: raw policy snapshots captured by the Open Terms Archive
genai-euproject across 11 GenAI providers (ChatGPT, Claude.ai, DeepSeek, Google Generative AI Services, Le Chat, Llama API, Meta AI, Microsoft Copilot, Perplexity, Qwen Chat, xAI). Unlike PGAv2, GenGA has no pre-tagged risk scores — all risk scoring and keyword tagging on these pages is LLM-assigned by direct reading, spanning 2025–present.
Note: This page contains documented policy clauses. Risk assessment is qualitative and context-dependent. For analysis of patterns across platforms, see:
Source: sources/GenGA/Claude.ai/Terms of Service/ (raw Markdown captures, Open Terms Archive genai-eu dataset — no pre-tagged JSONL exists for this source)
Last updated: 2025-11-11 (only capture in this dataset)
Note on methodology: No pre-computed
risk_score/keywordfields exist for this source. Keyword tags and clause analysis on this page are LLM-assigned, using the same rubric/vocabulary as the PGAv2 pages plus new GenAI-specific tags (input license,output ownership,competing model ban,feedback). Treat as first-pass analysis, not externally verified ground truth. This wiki’s risk-scoring system was retired project-wide on 2026-06-21 (seemethodology.md§4) — no page, GenGA included, computes or displays a numeric risk score.
Overview
1. What the document is: Anthropic’s consumer-facing Terms of Service for Claude.ai.
2. Input rights: “As between you and Anthropic… you retain any right, title, and interest that you have in such Inputs. Subject to your compliance with our Terms, we assign to you all our right, title, and interest (if any) in Outputs” — a user-protective ownership split, matching the pattern in Anthropic’s Commercial Terms. Separately, Anthropic claims a broad right to use Materials (Inputs + Outputs + Feedback) to “provide, maintain, and improve the Services… including training our models, unless you opt out of training through your account settings” — but two carve-outs, the same ones found in the Privacy Policy, override the opt-out: Feedback-rated conversations and safety-flagged content still train the models no matter what you choose.
3. Output restrictions: §2 bans using the Services “To develop any products or services that compete with our Services, including to develop or train any artificial intelligence or machine learning algorithms or models or resell the Services” — a broad competing-model ban that explicitly names both AI and ML algorithms/models.
4. Non-explicit predatory clauses: §5 Feedback: “you agree that we may use the Feedback however we choose without any obligation or other payment to you” — broad, unconditional, and uncompensated, the same pattern as every other provider in this dataset. Also notable: an asymmetric “No assignment” clause — Customer’s rights under the Terms “may not be transferred or assigned… without our prior written consent, but may be assigned by us without restriction.” This assignable clause runs one way only: Anthropic can assign its own rights freely, but the user needs Anthropic’s consent to assign theirs — the same pattern previously found on Facebook’s Terms of Service in the PGAv2 wiki.
5. Regulatory references: No explicit GDPR or EU AI Act citation in the Terms of Service body itself, but the Governing Law clause is EEA/Switzerland-aware: “if you are a consumer within the European Economic Area or Switzerland, you may also file legal disputes in your local courts, based on your local laws.”
6. Regional variation: Governed by Irish law generally, with an EEA/Switzerland consumer carve-out preserving local-court/local-law rights — the same EEA/Switzerland distinction found in Anthropic’s Commercial Terms and DPA.
7. Key risk to users: The default opt-out (not opt-in) training model is the central risk, consistent with the Privacy Policy. The broad competing-model ban and the asymmetric assignment clause are secondary points worth tracking.
Flagged Keywords & Risks (LLM-assigned)
output ownership,input license,train AI/models— the user keeps rights to their Inputs and is assigned rights to the Outputs, but Anthropic still uses Materials (Inputs, Outputs, and Feedback) to train its models by default unless the user opts out — and safety-flagged or Feedback-rated content trains the models even if they do opt out. Why it matters: this is the central risk clause on this page, and it mirrors the Privacy Policy’s training clause.competing model ban— users can’t use the Services to develop or train a competing AI or ML model. Why it matters: broader wording than some competitors’ equivalent clauses — it explicitly names both “artificial intelligence” and “machine learning” algorithms/models.feedback— Anthropic can use any feedback a user gives however it wants, with no restriction and no payment. Why it matters: identical pattern to every other provider’s Feedback clause in this dataset.assignable— “may not be transferred or assigned by you… but may be assigned by us without restriction.” Why it matters: this assignment clause runs one way only — Anthropic can assign its own rights freely, but the user needs Anthropic’s written consent to assign theirs. The same pattern was previously documented on Facebook’s Terms of Service in the PGAv2 wiki.
Regulatory & Research Context
Pandit et al. (2026) note that Anthropic’s Claude, among the six GenAI services they coded, used inputs and outputs “for other purposes beyond” the immediate service; the default opt-out (not opt-in) training clause here — with safety-flagged and Feedback-rated content always eligible regardless of opt-out — is a direct instance of that pattern. Davidson et al. (2026), who studied Anthropic directly, would situate the asymmetric “No assignment” clause (Anthropic’s rights freely assignable, the user’s restricted) and the broad “artificial intelligence or machine learning algorithms or models” competing-product ban within the kind of one-sided contractual structuring their comparative analysis of LLM Terms of Service documents across providers.
Changes Summary
| Date | What changed |
|---|---|
| 2025-11-11 | Baseline and only version captured in this dataset. |
Version History
2025-11-11
- Explicit AI clause: YES
- Non-explicit predatory: YES
- Flagged keywords:
output ownership,input license,train AI/models,competing model ban,feedback,assignable
Clause: output ownership, input license
Rights and Responsibilities. You are responsible for all Inputs you submit to our Services and all Actions. You must ensure that you have all rights, licenses, and permissions that are necessary for us to process such Inputs in accordance with our Terms and to provide the Services to you, including for example, to integrate with third-party services, to share Materials with others at your direction, and to take Actions. You must ensure that your submission of Inputs to us, your sharing them with others, or directing Claude to take Actions will not violate our Terms, our Acceptable Use Policy, or any laws or regulations applicable to those Inputs. As between you and Anthropic, and to the extent permitted by applicable law, you retain any right, title, and interest that you have in such Inputs. Subject to your compliance with our Terms, we assign to you all our right, title, and interest (if any) in Outputs.
Clause: train AI/models
Our use of Materials. We may use Materials to provide, maintain, and improve the Services and to develop other products and services, including training our models, unless you opt out of training through your account settings. Even if you opt out, we will use Materials for model training when: (1) you provide Feedback to us regarding any Materials, or (2) your Materials are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance our safety research.
Clause: competing model ban
To develop any products or services that compete with our Services, including to develop or train any artificial intelligence or machine learning algorithms or models or resell the Services.
Clause: feedback
We appreciate feedback, including ideas and suggestions for improvement or rating an Output in response to an Input (“Feedback”). If you rate an Output in response to an Input—for example, by using the thumbs up/thumbs down icon—we will store the related conversation as part of your Feedback. You have no obligation to give us Feedback, but if you do, you agree that we may use the Feedback however we choose without any obligation or other payment to you.
Clause: assignable
No assignment. These Terms may not be transferred or assigned by you without our prior written consent, but may be assigned by us without restriction.