Ethics Report: Alibaba
Rubric: Organisation v4 · Reviewed 3/22/2026
Little to no verifiable ethical commitment
Safety & Harm Reduction
6/25Dedicated safety / responsible-use policy
Publishes a dedicated safety/responsible-use policy that is publicly accessible.
Evidence
Alibaba has published safety-related policies and frameworks integrated into its regulatory compliance approach, including adherence to China's cybersecurity and data protection laws. In April 2021, the company established a technology ethics committee. However, no dedicated standalone safety/responsible-use policy page with specific enforceable terms and defined prohibited uses is evident from public sources. The safety measures described are primarily regulatory compliance-focused and embedded within broader operational frameworks rather than presented as a dedicated policy.
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 or red-team program is documented in available sources. While Alibaba maintains internal red-teaming platforms (Moyu and RedChain) for pre-deployment testing of its own models, these are not public programs and no documented results or findings from completed rounds are published. No HackerOne listing, public red-team call, or equivalent public program disclosure exists in the provided sources.
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
Alibaba published a joint report with the China Electronics Standardization Institute (CESI) in April 2025 detailing its approach to LLM safety, including implementation of SFT, DPO, Constitutional AI, and content filtering measures. However, this report contains limited quantitative safety data and does not provide comprehensive quantitative safety benchmarks across multiple harm categories. The report focuses primarily on technical methodology rather than measurable safety evaluation results. No third-party audit is mentioned.
Sources
Documented content-filtering / guardrails
Documents content-filtering/guardrails on production endpoints with user-facing documentation.
Evidence
Alibaba's content-filtering mechanisms are mentioned and described in its April 2025 CESI report. The company uses Constitutional AI, standardized Q&A libraries, and an input-output screening system with a keyword interception database covering at least 10,000 blocked terms across 31 risk categories. However, documentation focuses on existence of filters rather than detailed explanation of what specifically 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 dedicated reporting mechanisms or response timelines is evident in public sources. While Alibaba maintains internal security telemetry and detected the March 2024 autonomous agent incident through 'production security telemetry,' no public incident-response process, abuse reporting form, security email, or SLA documentation is disclosed. The sources indicate incident detection occurred but do not describe a public-facing incident-response mechanism.
Sources
Transparency & Trust
4/25Training data provenance disclosure
Publishes training data provenance disclosures identifying sources/types/datasets. "Publicly available data" alone = 0.
Evidence
Alibaba has disclosed general categories of training data sources. The article states that 'datasets are curated to reflect Chinese regulatory and cultural contexts' and that the company utilizes data aligned with national standards. However, no specific dataset names, sources, meaningful composition details, or filtering/exclusion criteria are publicly disclosed. The disclosure remains at the level of general categories rather than substantive detail about specific datasets or provenance.
Sources
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
Alibaba has published technical documentation describing aspects of its Qwen models, including model architecture families and the April 2025 CESI joint report detailing training approaches (SFT, DPO, Constitutional AI). However, documentation lacks comprehensive coverage of architecture details, scale specifications, full training approaches, and limitation disclosures. The technical documentation is partial rather than substantive in the manner required for a score of 5.
Sources
Transparency report (takedowns, government requests, etc.)
Publishes a transparency report covering takedowns, government requests, enforcement stats, and/or safety incidents.
Evidence
No transparency report disclosing takedowns, government requests, content removal statistics, or similar information is documented in available sources. Alibaba has published safety-related reports and registered AI products with the CAC, but no transparency report in the sense of disclosed government requests or content moderation actions is evident. The absence of such a report results in a score of 0.
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
No evidence of explicit disclosure regarding training data use of user/creator content in ToS, nor any opt-out mechanism is documented in public sources. While Alibaba operates under compliance frameworks and maintains safety measures, specific ToS language disclosing whether user data is used for training and any opt-out mechanism is not disclosed in available sources. The criterion requires explicit ToS language with either opt-out or commitment not to use data.
Sources
Creator/artist content provenance disclosure
Discloses training data provenance specifically for creator/artist content (copyrighted or artist-created works).
Evidence
No disclosure regarding creator/artist content provenance is documented. The sources do not indicate that Alibaba has disclosed content types, sources, licensing arrangements, or acknowledgment regarding the use or non-use of creative/copyrighted works in training data. Unlike general data provenance, specific disclosures about creative content are absent.
Sources
Human & Creator Impact
0/25Artist/creator opt-out or removal mechanism
Documented artist/creator opt-out or removal mechanism. "We respect copyright" alone = 0.
Evidence
No artist/creator opt-out mechanism or removal process is documented in available sources. While Alibaba publishes open-source models through ModelScope, no opt-out form, removal process (Spawning integration, email removal, etc.), or published evidence of honoring creator requests is disclosed. The open-source strategy does not constitute an opt-out mechanism for creators.
Sources
Public licensing or revenue-sharing with creators
Public licensing agreements or revenue-sharing partnerships with creators/publishers/media organizations.
Evidence
No public licensing or revenue-sharing deals with creators or artists are documented in the sources. While Alibaba has announced investments in AI startups like Moonshot AI and MiniMax, and maintains partnerships with organizations like Renmin University and Yara Asia, none of these represent licensing deals or creator compensation programs. No publicly announced partnership specifically involving creator or artist compensation is evident.
Sources
Provenance/attribution tooling for AI outputs
Provenance/attribution tooling for AI-generated outputs (C2PA, watermarking, metadata tagging).
Evidence
No tooling for provenance or attribution of AI outputs is documented, nor is a public commitment to any provenance standard (C2PA, watermarking, SynthID, etc.) disclosed in available sources. While Alibaba develops sophisticated technical safety measures, no specific commitment to or implementation of provenance/attribution tooling for outputs is mentioned.
Sources
Workforce impact assessment or commitment
Published workforce impact assessment or commitment (labor market effects, reskilling, human-in-the-loop programs).
Evidence
No workforce impact assessment or commitment to workforce transition support is documented in public sources. While the AI Safety Governance Framework 2.0 (which Alibaba contributed to) identifies risks concerning labor value diminishment and recommends human-in-the-loop systems, Alibaba itself has not published a specific statement naming an initiative, partner, or program addressing workforce impact. The contribution to policy development does not constitute an organizational workforce commitment or assessment.
Sources
Does NOT claim ownership over user-generated outputs
ToS does NOT claim copyright/exclusive ownership over user-generated outputs. Silent ToS = 0.
Evidence
No disclosure regarding user ownership of outputs is documented in available sources. While Alibaba provides AI services and tools, no explicit ToS statement is cited granting users full ownership, unrestricted rights, or clarifying the company's claims over user-generated outputs. The sources do not provide sufficient detail about output ownership terms in Alibaba's ToS to determine whether full user ownership is granted.
Sources
Governance
12/25Discloses corporate structure, investors, and board
Publicly discloses corporate structure, major investors, and board composition.
Evidence
Alibaba has publicly disclosed board information through its corporate structure documentation. The leadership is identified: Eddie Wu serves as CEO of both Alibaba Group and Cloud Intelligence Group, and Joseph Tsai serves as Chairman following the 2023 restructuring. However, the sources do not provide comprehensive disclosure of the full board membership or major investor lists through official channels. Only partial disclosure of top leadership is evident rather than both board and investors being fully disclosed.
Sources
Independent ethics/safety advisory board
Independent ethics/safety advisory board with verifiably external members. Internal trust & safety team alone = 0.
Evidence
Alibaba established a technology ethics committee in April 2021 led by the Group's Chief Technology Officer. The committee includes seven independent experts from legal, philosophical, and technological backgrounds, alongside members of the DAMO Academy and legal teams. However, documentation of the committee's independence is limited, no formal published mandate is evident, and no published recommendations or reports from the committee have been disclosed in available sources. This constitutes a named body with at least one verifiably external/independent member but lacks full independence documentation.
Sources
Legal corporate structure preserving safety/mission
Corporate structure preserves safety/mission mandate via a legal mechanism (PBC, capped-profit, charter clause).
Evidence
No verifiable legal mechanism preserving safety/mission in corporate filings is documented. The sources describe Alibaba's 2023 restructuring into a decentralized holding company with six business groups, but no evidence of special legal structures such as Benefit Corporation status, capped-profit provisions, or other legal mechanisms specifically designed to preserve safety/mission is disclosed. The restructuring was designed for operational decentralization and independent financing rather than safety preservation.
Sources
Public policy engagement or lobbying disclosure
Public policy engagement or lobbying disclosure: positions on AI regulation, lobbying spend, governance framework signatory.
Evidence
Alibaba has engaged with public policy and safety standards frameworks. The company was a key stakeholder in the development of China's AI Safety Governance Framework 2.0 (released September 2025) and participates in the Technical Committee 260 (TC260) for drafting AI risk standards. The company has also collaborated with state-affiliated bodies like the China Electronics Standardization Institute (CESI). However, no comprehensive lobbying disclosure or active engagement with international frameworks is documented. The engagement is primarily focused on Chinese domestic standards.
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 (VP-level or above) citing safety or ethics concerns are documented in available public sources. While Alibaba underwent leadership transitions (Daniel Zhang stepped down, replaced by Joseph Tsai and Eddie Wu), these were routine organizational restructuring moves unrelated to safety/ethics concerns. No on-record statements from departing executives citing safety or ethics reasons have been disclosed. The absence of such documented departures results in a clean record.
Sources
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.