Ethics Report: Perplexity
Rubric: Organisation v4 · Reviewed 3/22/2026
Little to no verifiable ethical commitment
Safety & Harm Reduction
2/25Dedicated safety / responsible-use policy
Publishes a dedicated safety/responsible-use policy that is publicly accessible.
Evidence
Perplexity has published general safety and ethics information focused primarily on hallucination mitigation through RAG architecture, evidence-first methodology, and structured citations. However, no dedicated safety/responsible-use policy page with specific, enforceable terms and defined prohibited uses is evident in public documentation. The article notes that 'specific public disclosures regarding the internal governance of the organization's safety teams or exhaustive lists of prohibited topics are less detailed in available technical documentation.'
Sources
Public bug-bounty or red-team program
Operates or funds a public bug-bounty / red-team program. Internal-only programs score 0.
Evidence
No public bug-bounty program or red-team initiative is documented in the provided research sources or article content. There is no mention of HackerOne listing, public red-team calls, or any formal vulnerability disclosure program.
Sources
Published safety evaluation within last 24 months
Published safety evaluation/audit/model card within last 24 months with quantitative benchmarks on harmful outputs (bias, toxicity, hallucination, etc.).
Evidence
No published safety evaluation, model card, or comprehensive safety benchmarks are evident in the provided research materials. While the article discusses architectural approaches to reducing hallucinations through RAG, there are no quantitative safety evaluations across multiple harm categories, nor third-party audits documented.
Sources
Documented content-filtering / guardrails
Documents content-filtering/guardrails on production endpoints with user-facing documentation.
Evidence
The article does not document specific content filtering or guardrails implementation. While it discusses RAG architecture and retrieval-based safeguards, there is no public documentation explaining what content is filtered, why, or how users can report false positives.
Sources
Documented incident-response process
Documented incident-response process for safety failures with a reporting mechanism. Generic "contact us" alone = 0.
Evidence
No documented incident-response process with defined reporting mechanisms or SLAs is presented in available documentation. No dedicated abuse reporting form, security email, or response timeline commitments are disclosed.
Sources
Transparency & Trust
6/25Training data provenance disclosure
Publishes training data provenance disclosures identifying sources/types/datasets. "Publicly available data" alone = 0.
Evidence
Perplexity discloses that its RAG architecture uses web data combined with LLM reasoning, and the article mentions 'retrieval of web indices or trusted knowledge repositories.' However, specific dataset names, sources, composition details, or data filtering/exclusion criteria are not comprehensively disclosed. The disclosure remains at the general categories level (web data, knowledge bases) without substantive curation detail.
Meaningful technical documentation for flagship model(s)
Publishes meaningful technical documentation (system card, tech report, research paper) for flagship model(s) regardless of whether weights are released.
Evidence
Basic technical documentation exists for the Sonar model family (noted as based on Llama 3.1 70B) and API architecture. The Perplexity API documentation covers model options and search modes, but lacks comprehensive technical reports covering architecture in depth, training approaches, and detailed limitation disclosures. Documentation is primarily marketing-oriented with limited substantive technical depth.
Transparency report (takedowns, government requests, etc.)
Publishes a transparency report covering takedowns, government requests, enforcement stats, and/or safety incidents.
Evidence
No transparency report regarding takedowns, government requests, content removal, or other disclosure categories is evident in the research sources or article. The article does not mention any such report being published.
Sources
ToS training data use disclosure with opt-out
ToS explicitly states whether user inputs/outputs are used for training, with opt-out mechanism if applicable.
Evidence
The article and research sources do not disclose explicit ToS language regarding training data use or any user opt-out mechanism. The content discusses product features and partnerships but does not address how user queries or data are used for training, nor mention an opt-out provision.
Sources
Creator/artist content provenance disclosure
Discloses training data provenance specifically for creator/artist content (copyrighted or artist-created works).
Evidence
Perplexity has disclosed partnerships with Getty Images for image content (multi-year image partnership explicitly confirmed) and has entered into revenue-sharing agreements with publishers including Time, Fortune, and Der Spiegel to address copyright concerns. However, the disclosure lacks comprehensive detail on specific creative content types, all sources, or complete licensing arrangements. The disclosure exists but remains general in acknowledging creative works.
Sources
Human & Creator Impact
7/25Artist/creator opt-out or removal mechanism
Documented artist/creator opt-out or removal mechanism. "We respect copyright" alone = 0.
Evidence
The article documents significant legal challenges and cease-and-desist letters from The New York Times, Condé Nast, and others over unauthorized scraping, with no documented artist/creator opt-out or removal mechanism mentioned. The absence of an actual mechanism combined with ongoing copyright disputes indicates no effective opt-out system exists.
Sources
Public licensing or revenue-sharing with creators
Public licensing agreements or revenue-sharing partnerships with creators/publishers/media organizations.
Evidence
Perplexity has announced at least two public licensing deals: a multi-year image partnership with Getty Images and a revenue-sharing program launched in July 2024 with publishers including Time, Fortune, and Der Spiegel. These represent multiple discrete arrangements but not a structured ongoing creator compensation program with measurable scope.
Sources
Provenance/attribution tooling for AI outputs
Provenance/attribution tooling for AI-generated outputs (C2PA, watermarking, metadata tagging).
Evidence
No public commitment to provenance standards or production implementation of provenance tooling (C2PA metadata, watermarks, SynthID, etc.) is documented. While the platform emphasizes inline citations and source attribution, this does not constitute machine-readable provenance metadata or commitment to industry standards.
Sources
Workforce impact assessment or commitment
Published workforce impact assessment or commitment (labor market effects, reskilling, human-in-the-loop programs).
Evidence
No published workforce impact assessment, educational partnership initiative, or commitment addressing displacement of workers in journalism, research, or content creation sectors is documented in the provided sources. The article discusses societal impact on business models but not workforce transition support.
Sources
Does NOT claim ownership over user-generated outputs
ToS does NOT claim copyright/exclusive ownership over user-generated outputs. Silent ToS = 0.
Evidence
Perplexity's ToS and product terms grant users full ownership over outputs. The platform explicitly supports user generation of content through tools like Perplexity Pages and reports being used for academic and professional contexts. No claim of ownership or exclusive license retention over user-generated content is evident in documentation.
Sources
Governance
9/25Discloses corporate structure, investors, and board
Publicly discloses corporate structure, major investors, and board composition.
Evidence
The article documents that investors include New Enterprise Associates (NEA), NVIDIA, Jeff Bezos via Bezos Expeditions, IVP, and Daniel Gross. Major investors are publicly disclosed through funding announcements. However, a board of directors is not publicly disclosed in available documentation. One of the two components (investors) is disclosed.
Sources
Independent ethics/safety advisory board
Independent ethics/safety advisory board with verifiably external members. Internal trust & safety team alone = 0.
Evidence
The Perplexity Blog mentions the introduction of a 'Perplexity Health Advisory Board' as of March 19, 2026, indicating an advisory body exists. However, the independence of this board is unclear, member names are not provided in the research sources, and no published mandate, recommendations, or reports are documented. The body exists but independence and external verification are unclear.
Sources
Legal corporate structure preserving safety/mission
Corporate structure preserves safety/mission mandate via a legal mechanism (PBC, capped-profit, charter clause).
Evidence
No evidence of special legal structures (Benefit Corporation, capped-profit, mission-aligned governance mechanism) is documented. Perplexity operates as a standard venture-backed private company without disclosed legal mechanisms preserving safety or mission.
Sources
Public policy engagement or lobbying disclosure
Public policy engagement or lobbying disclosure: positions on AI regulation, lobbying spend, governance framework signatory.
Evidence
The article and research sources do not document public policy engagement, lobbying disclosure, signing onto frameworks, or published policy positions. The organization's focus appears product-centric with no mentioned participation in AI governance frameworks or policy bodies.
Sources
No senior departures citing safety/ethics (last 36 months)
No publicly documented senior leadership (VP+) departures or whistleblower events citing safety/ethics concerns in the last 36 months. Only on-record statements count.
Evidence
No senior departures citing safety or ethics concerns are documented in the article or research sources. The leadership team (Aravind Srinivas as CEO, Denis Yarats as CTO, Johnny Ho, Andy Konwinski, Dmitry Shevelenko as CBO) remains intact with no reported departures on ethical or safety grounds. Clean record on public record.
Sources
Sources
- 1
- 2
- 3
- 4
- 5
Scores are generated using the Amallo Ethics Rubric (Organisation v4) based on publicly verifiable information. Each criterion is scored against defined tiers — only exact tier values are valid. Evidence is sourced from official documentation, research papers, and independent analyses. Scores may change as new information becomes available.