1. What the document is: Perplexity's API Terms of Service ("Sonar by Perplexity and/or Agentic Research API"), governing business/developer Customer use of the API.
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.
Perplexity – Developer Terms
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/Perplexity/Developer Terms/ (raw Markdown captures, Open Terms Archive genai-eu dataset — no pre-tagged JSONL exists for this source)
Last updated: 2026-01-24
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 (train AI/models,feedback,competing model ban). 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: Perplexity’s API Terms of Service (“Sonar by Perplexity and/or Agentic Research API”), governing business/developer Customer use of the API.
2. Input rights: The Customer retains ownership of Input. Perplexity asserts no ownership of Output and assigns back any rights it might obtain by operation of law. Section 2.3.3 makes an unconditional no-training commitment: “Perplexity shall not use (or authorize third parties to use) Customer Content to train, retrain, fine-tune or otherwise improve any generative artificial intelligence models” — joining Llama API’s and Claude.ai’s Commercial Terms as the dataset’s third unconditional, no-exceptions training commitment.
3. Output restrictions: Output is explicitly stated as potentially non-unique: “Customer’s Output may be the same as, or similar to, content from the internet; and other users of the Services may receive Output that is similar to, or the same as, Customer’s Output.” The Customer waives any claim that its Output is unique.
4. Non-explicit predatory clauses: A broad, uncompensated Feedback assignment (§3.2): “Perplexity shall be free to use, implement and exploit in any manner, and Customer hereby assigns to Perplexity, all rights… without remuneration to Customer.” A separate “Usage Data” clause gives Perplexity ownership of de-identified/aggregated usage metadata “for any lawful purpose… without remuneration to Customer.” There’s also a competing-product suspension right: Perplexity may terminate a Customer Application’s license if it is “competitive with Perplexity.”
5. Regulatory references: None directly in this document. Privacy/GDPR compliance is incorporated by reference to the separate DPA (§6.1), and a HIPAA Business Associate Agreement requirement is disclosed for Protected Health Information (§6.2).
6. Regional variation: None found. Governing law is California, with disputes adjudicated in San Francisco County.
7. Key risk to users: Low-to-moderate. The unconditional no-training commitment is strongly user-protective, but the uncompensated Feedback and Usage Data clauses, plus the competing-product suspension right, are standard-pattern risks consistent with this dataset’s other API/developer agreements.
Flagged Keywords & Risks (LLM-assigned)
train AI/models— “Perplexity shall not use (or authorize third parties to use) Customer Content to train, retrain, fine-tune or otherwise improve any generative artificial intelligence models.” Why it matters: this is an unconditional commitment, joining Llama API and Claude.ai as the dataset’s clearest no-training guarantees.feedback— “Perplexity shall be free to use, implement and exploit in any manner, and Customer hereby assigns to Perplexity, all rights in and to any and all ideas, suggestions, recommendations and/or feedback… without remuneration to Customer.” Why it matters: this is a full rights assignment, not just a license — the broadest Feedback structure seen in this dataset.competing model ban— Perplexity may terminate a Customer Application’s license if it is “competitive with Perplexity.” Why it matters: this is enforced through a suspension right rather than a flat usage prohibition — a different mechanism from other providers’ competing-model bans.
Regulatory & Research Context
Perplexity was not among the providers Pandit et al. (2026) or Davidson et al. (2026) directly studied, but Pandit et al.’s general finding that GenAI terms impose obligations on users “without remuneration” applies by analogy to two clauses on this page: the Feedback assignment, under which Perplexity may “use, implement and exploit in any manner” all assigned Feedback “without remuneration to Customer,” and the structurally identical Usage Data clause. The unconditional train AI/models no-training commitment — “Perplexity shall not use… Customer Content to train, retrain, fine-tune or otherwise improve any generative artificial intelligence models” — is a genuinely protective counterpoint, the kind of clause Davidson et al.’s “regulatory gray area” framework would treat as reducing rather than creating uncertainty for legitimate developer use of the API.
Changes Summary
| Date | What changed |
|---|---|
| 2025-11-11 | Baseline version (“Last updated: May 23, 2025”). |
| 2026-01-24 | Cosmetic only: product rebrand reference (“Sonar by Perplexity and/or Agentic Research”), a Third-Party Models & Terms List URL correction, and pronoun/cross-reference fixes in the indemnification section. No change to Model Training, Feedback, or Usage Data clauses. |
Version History
2025-11-11
- Explicit AI clause: YES
- Non-explicit predatory: YES
- Flagged keywords:
train AI/models,feedback,competing model ban
Clause: train AI/models
Model Training. Perplexity shall not use (or authorize third parties to use) Customer Content to train, retrain, fine-tune or otherwise improve any generative artificial intelligence models.
Clause: feedback
Feedback. Perplexity shall be free to use, implement and exploit in any manner, and Customer hereby assigns to Perplexity, all rights in and to any and all ideas, suggestions, recommendations and/or feedback from Customer relating to the Perplexity IP without remuneration to Customer.
Clause: competing model ban
Suspension of Customer Applications. Without limitation of its other rights hereunder or at law, if Perplexity determines that a Customer Application is unlawful, unethical, inconsistent with the technical or security requirements set forth in the API Documentation, inconsistent with the AUP, interfering with the operation or integrity of the Services, competitive with Perplexity, or likely to damage Perplexity’s reputation or otherwise bring Perplexity into dispute, Perplexity may terminate the license under Section 2.1(ii) in respect of such Customer Application and deactivate the relevant API Key(s).
Clause: Usage Data (not separately concept-tagged)
As between Customer and Perplexity, Perplexity retains all right, title, and interest in and to such Usage Data. Perplexity may use Usage Data for any lawful purpose (e.g., to assess the relative popularity of the Services’ features or to improve categorization algorithms) without remuneration to Customer.
2026-01-24
All flagged clauses unchanged from 2025-11-11 — cosmetic-only changes (product rebrand, link correction, pronoun fixes).