{
  "name": "Maple open-weight AI policy record",
  "description": "A dated record of the laws, rules, licenses and official findings that decide how AI model weights are released, with one primary source per entry.",
  "url": "https://maple.ai/open-weights",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "attribution": "Source: Maple, Open-weight AI policy (maple.ai/open-weights), data through 2026-10-20.",
  "sources": "One primary source per entry: a Federal Register or Official Journal text, a government or legislature page, a license file, a code repository, a standards body, or the company’s own post.",
  "updatedThrough": "2026-10-20",
  "lastChecked": "2026-10-20",
  "entries": [
    {
      "id": "2026-10-20-dhs-deepseek-exclusion-order",
      "url": "https://maple.ai/open-weights#2026-10-20-dhs-deepseek-exclusion-order",
      "date": "2026-10-20",
      "date_precision": "day",
      "actor": "U.S. Department of Homeland Security",
      "entity": "deepseek",
      "category": "law",
      "category_label": "Laws and bills",
      "title": "DHS publishes DeepSeek exclusion and removal order",
      "summary": "Notice: DHS publishes an order excluding DeepSeek products from civilian procurement and requiring removal of DeepSeek products and third-party products developed using DeepSeek implementations.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The notice publishes an order by the Secretary of Homeland Security under the Federal Acquisition Supply Chain Security Act of 2018, 41 U.S.C. 1321-1328. It covers Hangzhou DeepSeek Artificial Intelligence Co., Ltd and affiliated entities as defined in 48 CFR 2.101. The signed order is reproduced as images in the notice’s appendix. The order in the appendix carries an original release date of July 24, 2026; the Federal Register notice publishing it is dated October 20, 2026.",
      "license": null,
      "source_url": "https://www.federalregister.gov/d/2026-21447",
      "source_label": "Federal Register, 91 FR 66346 (Docket DHS-2026-0628)",
      "source_type": "register",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2026-08-14-bis-diffusion-rescission-agenda",
      "url": "https://maple.ai/open-weights#2026-08-14-bis-diffusion-rescission-agenda",
      "date": "2026-08-14",
      "date_precision": "day",
      "actor": "Bureau of Industry and Security",
      "entity": "us-export-controls",
      "category": "export",
      "category_label": "Export controls",
      "title": "Agenda lists planned diffusion rule rescission",
      "summary": "Notice: The Unified Agenda says a planned interim final rule will formally rescind the January 2025 AI diffusion rule, which BIS has stopped enforcing.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "In the Bureau of Industry and Security regulatory-plan section, RIN 0694-AJ90, 'Implementation of the AI Action Plan Through Export Controls,' is described as an interim final rule that will rescind the January 2025 Framework for Artificial Intelligence Diffusion and establish a replacement framework. The notice does not publish the rescission itself.",
      "license": null,
      "source_url": "https://www.govinfo.gov/content/pkg/FR-2026-08-14/html/2026-16603.htm",
      "source_label": "Federal Register, Introduction to the Unified Agenda, 2026-16603",
      "source_type": "register",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2026-07-27-ai-act-digital-omnibus",
      "url": "https://maple.ai/open-weights#2026-07-27-ai-act-digital-omnibus",
      "date": "2026-07-27",
      "date_precision": "day",
      "actor": "European Parliament and Council",
      "entity": "eu-ai-act",
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Digital Omnibus amends AI Act dates",
      "summary": "Law: Regulation (EU) 2026/1744 sets high-risk AI Act dates of December 2, 2027 for Annex III and August 2, 2028 for Annex I.",
      "says": "In cases of observed or reported circumvention of the preventive measures or other safeguards, adequate corrective measures should be taken provided that such measures are reasonable, taking into account the specific AI system, including its release and distribution strategy (such as open-source releases).",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The act is dated July 8, 2026 and enters into force on the third day after Official Journal publication. It replaces Article 113, third paragraph, points (a) and (c), and adds point (d). It does not replace point (b), the August 2, 2025 date for Chapter V. The Commission news page states that the omnibus enters into force on July 27, 2026.",
      "license": null,
      "source_url": "https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ:L_202601744",
      "source_label": "Official Journal, Regulation (EU) 2026/1744",
      "source_type": "journal",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2026-07-14-bis-advanced-computing-enforcement",
      "url": "https://maple.ai/open-weights#2026-07-14-bis-advanced-computing-enforcement",
      "date": "2026-07-14",
      "date_precision": "day",
      "actor": "Bureau of Industry and Security",
      "entity": "us-export-controls",
      "category": "export",
      "category_label": "Export controls",
      "title": "BIS describes where chip licenses apply",
      "summary": "Rule: BIS states it is enforcing the worldwide advanced-computing license requirement only for specified country groups, citing its May 13, 2025 guidance.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The final rule, published July 14, 2026, and effective July 10, 2026, removes the United Arab Emirates from Country Groups D:3 and D:4 and adds it to Country Group A:5. It says BIS is only enforcing the section 742.6(a)(6)(iii)(A) license requirement for destinations in Country Groups D:1, D:4, or D:5, with stated exclusions, and for entities headquartered in or with an ultimate parent headquartered in Country Group D:5 or Macau.",
      "license": null,
      "source_url": "https://www.govinfo.gov/content/pkg/FR-2026-07-14/html/2026-14132.htm",
      "source_label": "Federal Register, 2026-14132",
      "source_type": "register",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2026-05-31-bis-diffusion-enforcement-guidance",
      "url": "https://maple.ai/open-weights#2026-05-31-bis-diffusion-enforcement-guidance",
      "date": "2026-05-31",
      "date_precision": "day",
      "actor": "Bureau of Industry and Security",
      "entity": "us-export-controls",
      "category": "export",
      "category_label": "Export controls",
      "title": "BIS guidance on diffusion rule enforcement",
      "summary": "Guidance: BIS says its non-enforcement of the AI Diffusion Rule's new requirements does not cover exports for D:5 or Macau-headquartered entities.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The guidance says that in May 2025 BIS announced it would not enforce the AI Diffusion Rule's new compliance requirements. It says a license is still required under section 742.6(a)(6)(iii)(A) for specified advanced computing items when the recipient is headquartered in, or has an ultimate parent headquartered in, Country Group D:5 or Macau.",
      "license": null,
      "source_url": "https://www.bis.gov/media/documents/bis-guidance-may-31-2026.pdf",
      "source_label": "BIS guidance, May 31, 2026",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2026-05-01-caisi-deepseek-v4-pro",
      "url": "https://maple.ai/open-weights#2026-05-01-caisi-deepseek-v4-pro",
      "date": "2026-05-01",
      "date_precision": "day",
      "actor": "Center for AI Standards and Innovation",
      "entity": "deepseek",
      "category": "safety",
      "category_label": "Safety findings",
      "title": "CAISI evaluation of DeepSeek V4 Pro",
      "summary": "Evaluation: CAISI evaluates the open-weight model DeepSeek V4 Pro and states its capabilities lag the frontier by about 8 months.",
      "says": "In April 2026, the Center for AI Standards and Innovation (CAISI) evaluated the open-weight AI model DeepSeek V4 Pro (\"DeepSeek V4\").",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The page says CAISI's evaluations, which include non-public benchmarks, indicate that DeepSeek V4 performs similarly to GPT-5, and that DeepSeek's self-reported results placed it nearer to Opus 4.6 and GPT-5.4. It says CAISI served the model from cloud-based H200 and B200 GPUs.",
      "license": null,
      "source_url": "https://www.nist.gov/news-events/news/2026/05/caisi-evaluation-deepseek-v4-pro",
      "source_label": "NIST, CAISI Evaluation of DeepSeek V4 Pro",
      "source_type": "agency",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2026-04-02-gemma-4-apache",
      "url": "https://maple.ai/open-weights#2026-04-02-gemma-4-apache",
      "date": "2026-04-02",
      "date_precision": "day",
      "actor": "Google",
      "entity": "gemma",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Gemma 4 under Apache 2.0",
      "summary": "Release: Gemma 4 model weights under the Apache 2.0 license, in E2B, E4B, 26B MoE, and 31B sizes.",
      "says": "That’s why Gemma 4 is released under a commercially permissive Apache 2.0 license.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The post says “We are releasing the Gemma 4 model weights” in E2B, E4B, 26B MoE, and 31B sizes. It also says “This open-source license provides a foundation” under the heading “An open-source license,” and: “They complement our Gemini models, giving developers the industry's most powerful combination of both open and proprietary tools.” The terms page, last modified April 1, 2026, still points models other than Gemma 4 to the Gemma Terms of Use.",
      "license": {
        "name": "Apache License, Version 2.0",
        "url": "https://ai.google.dev/gemma/apache_2",
        "commercial": "each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form.",
        "previous": {
          "label": "Gemma Terms of Use",
          "url": "https://ai.google.dev/gemma/terms"
        },
        "weights": true
      },
      "source_url": "https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/",
      "source_label": "Google blog, Gemma 4",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2026-03-27-new-york-raise-chapter-amendment",
      "url": "https://maple.ai/open-weights#2026-03-27-new-york-raise-chapter-amendment",
      "date": "2026-03-27",
      "date_precision": "day",
      "actor": "New York",
      "entity": null,
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Chapter amendment replaces the RAISE Act",
      "summary": "Law signed: Senate Bill S8828, chapter 96, repeals and replaces the RAISE Act text and defines frontier model and model weight.",
      "says": "CYBERSECURITY PRACTICES TO SECURE UNRELEASED MODEL WEIGHTS FROM UNAUTHORIZED MODIFICATION OR TRANSFER BY INTERNAL OR EXTERNAL PARTIES;",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The Senate actions list 'Mar 27, 2026' as 'signed chap.96.' The bill text defines a frontier model as a foundation model trained with more than 10^26 operations and a large frontier developer as one with annual gross revenue above $500 million. The sponsor memo calls it a chapter amendment to Chapter 699 of the Laws of 2025.",
      "license": null,
      "source_url": "https://www.nysenate.gov/legislation/bills/2025/S8828",
      "source_label": "New York Senate Bill S8828",
      "source_type": "legislature",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2025-12-19-new-york-raise-act-signed",
      "url": "https://maple.ai/open-weights#2025-12-19-new-york-raise-act-signed",
      "date": "2025-12-19",
      "date_precision": "day",
      "actor": "Governor Kathy Hochul",
      "entity": null,
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Hochul signs the New York RAISE Act",
      "summary": "Law signed: Governor Hochul signed the RAISE Act, S6953B/A6453B, with agreed chapter amendments requiring large frontier developers to publish safety protocols.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The release says the agreed-upon chapter amendments require large AI developers to publish information about their safety protocols and to report incidents within 72 hours, and create an oversight office in the Department of Financial Services. It states penalties of up to $1 million for a first violation and up to $3 million for subsequent violations.",
      "license": null,
      "source_url": "https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-require-ai-frameworks-ai-frontier-models",
      "source_label": "Governor Hochul press release, December 19, 2025",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-12-18-aisi-frontier-ai-trends",
      "url": "https://maple.ai/open-weights#2025-12-18-aisi-frontier-ai-trends",
      "date": "2025-12-18",
      "date_precision": "day",
      "actor": "AI Security Institute",
      "entity": null,
      "category": "safety",
      "category_label": "Safety findings",
      "title": "Report on open-weight safeguards",
      "summary": "Evaluation: The AI Security Institute's Frontier AI Trends Report gives findings on safeguards for open-weight systems, whose weights are directly accessible.",
      "says": "Open-weight AI systems – where the weights are directly accessible – are particularly hard to safeguard against misuse.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The report says basic techniques can remove trained-in refusal, other safeguard components can be disabled, and jailbreaks cannot be patched once the weights are no longer hosted by the defender. Its glossary defines an open-weight model as an AI model whose parameters are made freely available, and an open-source model as one whose parameters, code, and training data are made freely available.",
      "license": null,
      "source_url": "https://www.aisi.gov.uk/frontier-ai-trends-report",
      "source_label": "AI Security Institute, Frontier AI Trends Report",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-12-12-caisi-kimi-k2-thinking",
      "url": "https://maple.ai/open-weights#2025-12-12-caisi-kimi-k2-thinking",
      "date": "2025-12-12",
      "date_precision": "day",
      "actor": "Center for AI Standards and Innovation",
      "entity": null,
      "category": "safety",
      "category_label": "Safety findings",
      "title": "CAISI evaluation of Kimi K2 Thinking",
      "summary": "Evaluation: CAISI evaluates Kimi K2 Thinking, which it calls an open-weight model released November 6, 2025, and compares it with U.S. models.",
      "says": "Kimi K2 Thinking, an open-weight AI model released on November 6, 2025 by Moonshot AI.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The page says that at release Kimi K2 Thinking was the most capable model from a PRC-based developer in this evaluation and still lagged leading U.S. models, while its capabilities improved on the previous open-weight model frontier. It also says downloads a month after release were about 10 percent of DeepSeek R1's and under 5 percent of gpt-oss's.",
      "license": null,
      "source_url": "https://www.nist.gov/news-events/news/2025/12/caisi-evaluation-kimi-k2-thinking",
      "source_label": "NIST, CAISI Evaluation of Kimi K2 Thinking",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-12-05-gpai-training-summary-template",
      "url": "https://maple.ai/open-weights#2025-12-05-gpai-training-summary-template",
      "date": "2025-12-05",
      "date_precision": "day",
      "actor": "European Commission",
      "entity": "eu-ai-act",
      "category": "report",
      "category_label": "Reports and plans",
      "title": "Template for GPAI training summaries",
      "summary": "Communication: The Commission’s December 5, 2025 communication sets out an explanatory notice and template for public summaries of general-purpose AI model training content.",
      "says": "Providers of all general-purpose AI models placed on the Union market must fulfil the above obligation, including providers of general-purpose AI models released under free and open-source licenses, in so far as the models fall within the scope of the AI Act.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "A footnote on the same page says the Article 53(2) exception for a free and open-source general-purpose AI model does not apply to the obligation to make the summary public. Paragraph (1) says the Chapter V rules apply as of August 2, 2025.",
      "license": null,
      "source_url": "https://ec.europa.eu/newsroom/dae/redirection/document/118480",
      "source_label": "European Commission, training-summary template, C(2025) 8311",
      "source_type": "agency",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2025-11-19-gpai-provider-guidelines",
      "url": "https://maple.ai/open-weights#2025-11-19-gpai-provider-guidelines",
      "date": "2025-11-19",
      "date_precision": "day",
      "actor": "European Commission",
      "entity": "eu-ai-act",
      "category": "report",
      "category_label": "Reports and plans",
      "title": "Guidelines on GPAI provider obligations",
      "summary": "Report: The Commission’s guidelines for general-purpose AI model providers, C(2025) 7719 final, dated November 19, 2025, include an exception for models released under a free and open-source license.",
      "says": "These transparency-related obligations do not apply to providers of general-purpose AI models that are released under a free and open-source licence, under certain conditions, unless the model is a general-purpose AI model with systemic risk.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The guidelines say Article 53(2) and Article 54(6) except some duties for models released under a free and open-source license whose parameters, including the weights, are public, unless the model has systemic risk. Copyright policy and the training-content summary stay in place. The text also restates the August 2, 2025, August 2, 2026, and August 2, 2027 dates. The Commission’s library page lists the guidelines with “Publication 18 July 2025” and a Communication approving “the content of the draft Guidelines.”",
      "license": null,
      "source_url": "https://ec.europa.eu/newsroom/dae/redirection/document/118340",
      "source_label": "European Commission, GPAI guidelines, C(2025) 7719",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-10-29-gpt-oss-safeguard",
      "url": "https://maple.ai/open-weights#2025-10-29-gpt-oss-safeguard",
      "date": "2025-10-29",
      "date_precision": "day",
      "actor": "OpenAI",
      "entity": "gpt-oss",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "gpt-oss-safeguard research preview",
      "summary": "Release: gpt-oss-safeguard-120b and gpt-oss-safeguard-20b weights under the same Apache 2.0 license as gpt-oss.",
      "says": "Today, we’re releasing a research preview of gpt-oss-safeguard, our open-weight reasoning models for safety classification tasks, available in two sizes: gpt-oss-safeguard-120b and gpt-oss-safeguard-20b. These models are fine-tuned versions of our gpt-oss open models and available under the same permissive Apache 2.0 license, allowing anyone to use, modify, and deploy them freely. Both models can be downloaded today from Hugging Face.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The post describes the two models as fine-tuned versions of the gpt-oss models and says they are under the same Apache 2.0 license. The license file quoted here is the LICENSE in openai/gpt-oss, opened with the August 5, 2025 release.",
      "license": {
        "name": "Apache License, Version 2.0",
        "url": "https://github.com/openai/gpt-oss/blob/main/LICENSE",
        "commercial": "each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form.",
        "previous": null,
        "weights": true
      },
      "source_url": "https://openai.com/index/introducing-gpt-oss-safeguard/",
      "source_label": "OpenAI, Introducing gpt-oss-safeguard",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-09-30-caisi-deepseek-evaluation",
      "url": "https://maple.ai/open-weights#2025-09-30-caisi-deepseek-evaluation",
      "date": "2025-09-30",
      "date_precision": "day",
      "actor": "Center for AI Standards and Innovation",
      "entity": "deepseek",
      "category": "safety",
      "category_label": "Safety findings",
      "title": "CAISI evaluation of DeepSeek models",
      "summary": "Evaluation: CAISI reports on DeepSeek models it describes as open-weight, including jailbreaking results and a comparison with U.S. models.",
      "says": "To evaluate DeepSeek models, which are available as open-weight models, CAISI downloaded their model weights from the model sharing platform Hugging Face and deployed the models on CAISI's own cloud-based servers.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The report evaluates DeepSeek R1, R1-0528, and V3.1 against GPT-5, GPT-5-mini, gpt-oss, and Opus 4. It says DeepSeek models were much more likely to follow malicious hijacking instructions than the evaluated U.S. frontier models, and that the evaluated U.S. open-weight model matched or exceeded DeepSeek on that measure.",
      "license": null,
      "source_url": "https://www.nist.gov/system/files/documents/2025/09/30/CAISI_Evaluation_of_DeepSeek_AI_Models.pdf",
      "source_label": "CAISI, Evaluation of DeepSeek AI Models",
      "source_type": "agency",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2025-09-29-california-sb-53-chaptered",
      "url": "https://maple.ai/open-weights#2025-09-29-california-sb-53-chaptered",
      "date": "2025-09-29",
      "date_precision": "day",
      "actor": "California",
      "entity": "california",
      "category": "law",
      "category_label": "Laws and bills",
      "title": "California chapters frontier AI transparency act",
      "summary": "Law signed: Senate Bill 53 was approved and chaptered as Chapter 138, the Transparency in Frontier Artificial Intelligence Act.",
      "says": "Cybersecurity practices to secure unreleased model weights from unauthorized modification or transfer by internal or external parties.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The code says Chapter 25.1 was added by Stats. 2025, Ch. 138, Sec. 2 (SB 53), effective January 1, 2026. It defines a frontier model as a foundation model trained using more than 10^26 operations and defines 'model weight' as a numerical parameter adjusted through training.",
      "license": null,
      "source_url": "https://leginfo.legislature.ca.gov/faces/codes_displayText.xhtml?lawCode=BPC&division=8.&chapter=25.1.",
      "source_label": "California Business and Professions Code, Chapter 25.1",
      "source_type": "legislature",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2025-08-26-china-ai-plus-open-source",
      "url": "https://maple.ai/open-weights#2025-08-26-china-ai-plus-open-source",
      "date": "2025-08-26",
      "date_precision": "day",
      "actor": "State Council",
      "entity": null,
      "category": "report",
      "category_label": "Reports and plans",
      "title": "State Council on AI open source",
      "summary": "Plan: The State Council issues an opinion that supports AI open-source communities and opening models, tools, and datasets.",
      "says": "支持人工智能开源社区建设，促进模型、工具、数据集等汇聚开放，培育优质开源项目。",
      "says_kind": "quote",
      "says_lang": "zh",
      "details": "Guofa [2025] No. 11 is dated August 21, 2025 and was posted on August 26, 2025. A separate paragraph says China will promote making AI technology open-source and accessible in international cooperation. The opinion also says safety assessment and filing rules for AI should be improved.",
      "license": null,
      "source_url": "https://www.gov.cn/zhengce/content/202508/content_7037861.htm",
      "source_label": "State Council, Opinions on the AI Plus action, Guofa [2025] No. 11",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-08-05-gpt-oss",
      "url": "https://maple.ai/open-weights#2025-08-05-gpt-oss",
      "date": "2025-08-05",
      "date_precision": "day",
      "actor": "OpenAI",
      "entity": "gpt-oss",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "gpt-oss open-weight models",
      "summary": "Release: gpt-oss-120b and gpt-oss-20b weights under the Apache 2.0 license, described as open-weight language models.",
      "says": "We’re releasing gpt-oss-120b and gpt-oss-20b—two state-of-the-art open-weight language models that deliver strong real-world performance at low cost. Available under the flexible Apache 2.0 license",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The post says “gpt-oss models are our first open-weight language models since GPT-2” and that the weights are “freely available for download on Hugging Face.” On the same page: “Once an open-weight model is released, adversaries may be able to fine-tune the model for malicious purposes.” It says that even with robust fine-tuning, maliciously fine-tuned models “were unable to reach high capability levels” under the Preparedness Framework, and that “These findings informed our decision to release the gpt-oss models.”",
      "license": {
        "name": "Apache License, Version 2.0",
        "url": "https://github.com/openai/gpt-oss/blob/main/LICENSE",
        "commercial": "each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form.",
        "previous": null,
        "weights": true
      },
      "source_url": "https://openai.com/index/introducing-gpt-oss/",
      "source_label": "OpenAI, Introducing gpt-oss",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-07-23-ai-action-plan-open-models",
      "url": "https://maple.ai/open-weights#2025-07-23-ai-action-plan-open-models",
      "date": "2025-07-23",
      "date_precision": "day",
      "actor": "The White House",
      "entity": null,
      "category": "report",
      "category_label": "Reports and plans",
      "title": "Action Plan section on open-weight AI",
      "summary": "Plan: The AI Action Plan says open-source and open-weight models can be downloaded and modified, and that a release decision is up to the developer.",
      "says": "Open-source and open-weight AI models are made freely available by developers for anyone in the world to download and modify.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The section 'Encourage Open-Source and Open-Weight AI' also says the decision of whether and how to release an open or closed model is fundamentally up to the developer. A recommended action assigns the Commerce Department, through NTIA, to convene stakeholders on adoption by small and medium-sized businesses.",
      "license": null,
      "source_url": "https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf",
      "source_label": "Winning the AI Race: America's AI Action Plan",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-07-10-gpai-code-of-practice",
      "url": "https://maple.ai/open-weights#2025-07-10-gpai-code-of-practice",
      "date": "2025-07-10",
      "date_precision": "day",
      "actor": "European Commission",
      "entity": "eu-ai-act",
      "category": "report",
      "category_label": "Reports and plans",
      "title": "GPAI Code of Practice published",
      "summary": "Report: The Commission publishes a voluntary code of practice for providers of general-purpose AI models under the AI Act.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The page says the code was published on July 10, 2025, and that the Commission and the AI Board have confirmed it is an adequate voluntary tool. The Transparency and Copyright chapters address Article 53. The Safety and Security chapter addresses Article 55 models with systemic risk.",
      "license": null,
      "source_url": "https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai",
      "source_label": "European Commission, The General-Purpose AI Code of Practice",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-05-13-bis-diffusion-rule-nonenforcement",
      "url": "https://maple.ai/open-weights#2025-05-13-bis-diffusion-rule-nonenforcement",
      "date": "2025-05-13",
      "date_precision": "day",
      "actor": "Department of Commerce",
      "entity": "us-export-controls",
      "category": "export",
      "category_label": "Export controls",
      "title": "Commerce directs non-enforcement of diffusion rule",
      "summary": "Announcement: Commerce instructs staff not to enforce the AI Diffusion Rule and says BIS plans a regulation formalizing rescission.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The release, headed for immediate release on May 13, 2025, says the AI Diffusion Rule was issued on January 15, 2025, with compliance requirements set for May 15, 2025. It says Under Secretary Jeffery Kessler instructed enforcement officials not to enforce the rule, and that BIS plans to publish a regulation formalizing the rescission and a replacement rule.",
      "license": null,
      "source_url": "https://www.bis.gov/press-release/department-commerce-announces-rescission-biden-era-artificial-intelligence-diffusion-rule-strengthens",
      "source_label": "Department of Commerce press release, May 13, 2025",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-04-29-qwen3",
      "url": "https://maple.ai/open-weights#2025-04-29-qwen3",
      "date": "2025-04-29",
      "date_precision": "day",
      "actor": "Qwen Team",
      "entity": "qwen",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Qwen3 weights under Apache 2.0",
      "summary": "Release: Qwen3 MoE and dense model weights under the Apache 2.0 license, including Qwen3-235B-A22B.",
      "says": "We are open-weighting two MoE models: Qwen3-235B-A22B, a large model with 235 billion total parameters and 22 billion activated parameters, and Qwen3-30B-A3B, a smaller MoE model with 30 billion total parameters and 3 billion activated parameters. Additionally, six dense models are also open-weighted, including Qwen3-32B, Qwen3-14B, Qwen3-8B, Qwen3-4B, Qwen3-1.7B, and Qwen3-0.6B, under Apache 2.0 license.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The post lists Qwen3-235B-A22B among the models it says are open-weighted under the Apache 2.0 license. On June 7, 2024, Qwen2-72B had remained on the Qianwen License. The LICENSE file opened for this quote is on Qwen/Qwen3-8B. Its appendix reads “Copyright 2024 Alibaba Cloud” and “Licensed under the Apache License, Version 2.0.”",
      "license": {
        "name": "Apache License, Version 2.0",
        "url": "https://huggingface.co/Qwen/Qwen3-8B/blob/main/LICENSE",
        "commercial": "each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form.",
        "previous": {
          "label": "Tongyi Qianwen LICENSE AGREEMENT",
          "url": "https://huggingface.co/Qwen/Qwen-7B/blob/main/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://qwenlm.github.io/blog/qwen3/",
      "source_label": "Qwen blog, Qwen3",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-04-05-llama-4",
      "url": "https://maple.ai/open-weights#2025-04-05-llama-4",
      "date": "2025-04-05",
      "date_precision": "day",
      "actor": "Meta",
      "entity": "llama",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Llama 4 Scout and Maverick",
      "summary": "Release: Llama 4 Scout and Maverick weights for download; the same post says it is not yet releasing Llama 4 Behemoth.",
      "says": "We’re introducing Llama 4 Scout and Llama 4 Maverick, the first open-weight natively multimodal models with unprecedented context length support and our first built using a mixture-of-experts (MoE) architecture.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The post says Scout and Maverick are “available for download today” on llama.com and Hugging Face, and: “While we’re not yet releasing Llama 4 Behemoth as it is still training, we’re excited to share more technical details about our approach.” It describes Behemoth as 288B active parameters, 16 experts, and nearly two trillion total parameters. The license states “Llama 4 Version Effective Date: April 5, 2025,” and the 700 million clause is keyed to “the Llama 4 version release date.” The acceptable-use policy is https://www.llama.com/llama4/use-policy.",
      "license": {
        "name": "LLAMA 4 COMMUNITY LICENSE AGREEMENT",
        "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama4/LICENSE",
        "commercial": "If, on the Llama 4 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.",
        "previous": {
          "label": "LLAMA 3.3 COMMUNITY LICENSE AGREEMENT",
          "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://ai.meta.com/blog/llama-4-multimodal-intelligence/",
      "source_label": "Meta AI blog, Llama 4",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-03-14-china-ai-content-labels",
      "url": "https://maple.ai/open-weights#2025-03-14-china-ai-content-labels",
      "date": "2025-03-14",
      "date_precision": "day",
      "actor": "Cyberspace Administration of China",
      "entity": null,
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Labels required on AI-generated content",
      "summary": "Rule: Four Chinese agencies issue measures on labels for AI-generated synthetic content, effective September 1, 2025.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The notice is dated March 7, 2025 and posted March 14, 2025. The measures apply to network information service providers in the situations covered by the algorithm-recommendation rules, the deep-synthesis rules, and the generative AI service measures. They require explicit and implicit labels on generated or synthetic text, images, audio, video, and virtual scenes.",
      "license": null,
      "source_url": "https://www.cac.gov.cn/2025-03/14/c_1743654684782215.htm",
      "source_label": "CAC, Measures for Labeling AI-Generated Content",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-02-12-france-india-open-models",
      "url": "https://maple.ai/open-weights#2025-02-12-france-india-open-models",
      "date": "2025-02-12",
      "date_precision": "day",
      "actor": "France and India",
      "entity": null,
      "category": "report",
      "category_label": "Reports and plans",
      "title": "France and India on open models",
      "summary": "Statement: France and India say they will pursue research on large language models that are open and freely reusable.",
      "says": "pour approfondir les partenariats de recherche pour développer des modèles de langage larges, ouverts et librement réutilisables afin de soutenir et promouvoir la diversité linguistique",
      "says_kind": "quote",
      "says_lang": "fr",
      "details": "The same declaration says the two countries will support shared resources including open-source tools, and that a framework should include free and open resources for countries, innovators, researchers, and citizens. The page is the French text of the joint declaration.",
      "license": null,
      "source_url": "https://www.elysee.fr/emmanuel-macron/2025/02/12/declaration-franco-indienne-sur-lintelligence-artificielle",
      "source_label": "Élysée, Franco-Indian declaration on artificial intelligence",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2025-02-06-ostp-ai-action-plan-rfi",
      "url": "https://maple.ai/open-weights#2025-02-06-ostp-ai-action-plan-rfi",
      "date": "2025-02-06",
      "date_precision": "day",
      "actor": "Networking and Information Technology Research and Development (NITRD) National Coordination Office (NCO), National Science Foundation",
      "entity": null,
      "category": "report",
      "category_label": "Reports and plans",
      "title": "OSTP request lists open source development",
      "summary": "Request for information: On behalf of OSTP, NITRD asks for AI Action Plan comments and lists open source development among the topics.",
      "says": "Responses can address any relevant AI policy topic, including but not limited to: hardware and chips, data centers, energy consumption and efficiency, model development, open source development, application and use",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The notice was published February 6, 2025, by the NITRD National Coordination Office at the National Science Foundation, on behalf of OSTP. Comments were due March 15, 2025. It cites Executive Order 14179 of January 23, 2025, and says Executive Order 14148 of January 20, 2025, revoked Executive Order 14110.",
      "license": null,
      "source_url": "https://www.govinfo.gov/content/pkg/FR-2025-02-06/html/2025-02305.htm",
      "source_label": "Federal Register, 90 FR 9088",
      "source_type": "register",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2025-01-20-deepseek-r1-mit",
      "url": "https://maple.ai/open-weights#2025-01-20-deepseek-r1-mit",
      "date": "2025-01-20",
      "date_precision": "day",
      "actor": "DeepSeek",
      "entity": "deepseek",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "DeepSeek-R1 moves to MIT",
      "summary": "License change: DeepSeek-R1 code and models are released under the MIT License.",
      "says": "Code and models are released under the MIT License: Distill & commercialize freely!",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The same news page says “Fully open-source model & technical report,” “DeepSeek-R1 is now MIT licensed for clear open access,” “Open for the community to leverage model weights & outputs,” and “6 small models fully open-sourced.” The MIT file on the R1 repository begins “Copyright (c) 2023 DeepSeek.” The DeepSeek-V3 LICENSE-MODEL file remained the October 23, 2023 DeepSeek License Agreement when fetched on October 16, 2026.",
      "license": {
        "name": "MIT License",
        "url": "https://huggingface.co/deepseek-ai/DeepSeek-R1/blob/main/LICENSE",
        "commercial": "Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the \"Software\"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software",
        "previous": {
          "label": "DEEPSEEK LICENSE AGREEMENT",
          "url": "https://github.com/deepseek-ai/DeepSeek-LLM/blob/main/LICENSE-MODEL"
        },
        "weights": true
      },
      "source_url": "https://api-docs.deepseek.com/news/news250120/",
      "source_label": "DeepSeek API docs news, 250120",
      "source_type": "company",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2025-01-15-bis-eccn-4e091-diffusion-rule",
      "url": "https://maple.ai/open-weights#2025-01-15-bis-eccn-4e091-diffusion-rule",
      "date": "2025-01-15",
      "date_precision": "day",
      "actor": "Bureau of Industry and Security",
      "entity": "us-export-controls",
      "category": "export",
      "category_label": "Export controls",
      "title": "BIS controls some closed model weights",
      "summary": "Rule: BIS adds ECCN 4E091 for certain closed-weight model weights and states it is not imposing controls on open-weight models.",
      "says": "Accordingly, BIS and its interagency partners are not today imposing controls on open-weight models.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The interim final rule, 90 FR 4544, is effective January 13, 2025, with compliance for most changes on May 15, 2025. Its text says ECCN 4E091 does not control the parameters of a model that has been published as defined in section 734.7(a) of the EAR, and that no open-weight model was known to have been trained on more than 10^26 operations.",
      "license": null,
      "source_url": "https://www.govinfo.gov/content/pkg/FR-2025-01-15/html/2025-00636.htm",
      "source_label": "Federal Register, Framework for Artificial Intelligence Diffusion, 90 FR 4544",
      "source_type": "register",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-12-25-deepseek-v3",
      "url": "https://maple.ai/open-weights#2024-12-25-deepseek-v3",
      "date": "2024-12-26",
      "date_precision": "day",
      "actor": "DeepSeek",
      "entity": "deepseek",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "DeepSeek-V3 model license",
      "summary": "Release: DeepSeek-V3 weights in the Hugging Face repository under the DeepSeek License Agreement, Version 1.0, dated October 23, 2023.",
      "says": "Fully open-source models & papers",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The file on the DeepSeek-V3 repository, fetched October 16, 2026, is still the DeepSeek License Agreement, Version 1.0, October 23, 2023, the same text as on DeepSeek-LLM. It is not the MIT License. The repository lists 163 safetensors shards. Use-based restrictions in Attachment A must be passed downstream, and section 7 reserves a right to restrict usage that violates the license.",
      "license": {
        "name": "DEEPSEEK LICENSE AGREEMENT",
        "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3/blob/main/LICENSE-MODEL",
        "commercial": "a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare, publicly display, publicly perform, sublicense, and distribute the Complementary Material, the Model, and Derivatives of the Model.",
        "previous": null,
        "weights": true
      },
      "source_url": "https://api-docs.deepseek.com/news/news1226/",
      "source_label": "DeepSeek API docs, Introducing DeepSeek-V3",
      "source_type": "company",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2024-12-17-model-openness-framework",
      "url": "https://maple.ai/open-weights#2024-12-17-model-openness-framework",
      "date": "2024-12-17",
      "date_precision": "day",
      "actor": "LF AI & Data",
      "entity": null,
      "category": "definition",
      "category_label": "Definitions and standards",
      "title": "Model Openness Framework 1.0",
      "summary": "Definition: Model Openness Framework 1.0 defines Class I, Class II, and Class III, and defines an open license.",
      "says": "An open license is a license allowing unrestricted usage, study, modification, and redistribution for any purpose.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The specification is titled as Version 1.0, dated Dec 17, 2024, from LF AI & Data Generative AI Commons, and it says “This is a stable document.” It says: “The MOF defines three classes (Class III: Open Model, Class II: Open Tooling Model, and Class I: Open Science Model).” Class III includes model architecture and final model parameters under type-appropriate open licenses. For model parameters it lists preferred CDLA-Permissive-2.0 and acceptable OSI-approved or open data licenses.",
      "license": null,
      "source_url": "https://lfaidata.foundation/wp-content/uploads/sites/3/2025/01/05_White_paper_MOF_Specification.pdf",
      "source_label": "LF AI & Data, Model Openness Framework Specification",
      "source_type": "standards",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-12-06-llama-3-3",
      "url": "https://maple.ai/open-weights#2024-12-06-llama-3-3",
      "date": "2024-12-06",
      "date_precision": "day",
      "actor": "Meta",
      "entity": "llama",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Llama 3.3 community license",
      "summary": "Release: Llama 3.3 weights under the Llama 3.3 Community License Agreement, which includes a 700 million monthly active user clause.",
      "says": "including machine-learning model code, trained model weights, inference-enabling code, training-enabling code, fine-tuning enabling code",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The license states “Llama 3.3 Version Release Date: December 6, 2024.” The 700 million monthly active user clause remains. The acceptable-use policy linked from the license is https://www.llama.com/llama3_3/use-policy. The llama-models README table lists Llama 3.3 as 12/04/2024.",
      "license": {
        "name": "LLAMA 3.3 COMMUNITY LICENSE AGREEMENT",
        "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/LICENSE",
        "commercial": "If, on the Llama 3.3 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.",
        "previous": {
          "label": "LLAMA 3.2 COMMUNITY LICENSE AGREEMENT",
          "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/LICENSE",
      "source_label": "llama-models LICENSE, Llama 3.3",
      "source_type": "license",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2024-10-28-osi-open-source-ai-definition",
      "url": "https://maple.ai/open-weights#2024-10-28-osi-open-source-ai-definition",
      "date": "2024-10-28",
      "date_precision": "day",
      "actor": "Open Source Initiative",
      "entity": null,
      "category": "definition",
      "category_label": "Definitions and standards",
      "title": "Open Source AI Definition 1.0",
      "summary": "Definition: The Open Source AI Definition 1.0 requires parameters under OSI-approved terms, plus data information and code.",
      "says": "An Open Source AI is an AI system made available under terms and in a way that grant the freedoms to: Use the system for any purpose and without having to ask for permission. Study how the system works and inspect its components. Modify the system for any purpose, including to change its output. Share the system for others to use with or without modifications, for any purpose.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The definition page says the preferred form of making modifications includes data information, code, and parameters, and that “Parameters shall be made available under OSI-approved terms.” It also says: “Open Source models and Open Source weights must include the data information and code used to derive those parameters.”",
      "license": null,
      "source_url": "https://opensource.org/ai/open-source-ai-definition",
      "source_label": "Open Source Initiative, Open Source AI Definition",
      "source_type": "standards",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2024-09-29-california-sb-1047-veto",
      "url": "https://maple.ai/open-weights#2024-09-29-california-sb-1047-veto",
      "date": "2024-09-29",
      "date_precision": "day",
      "actor": "Governor Gavin Newsom",
      "entity": "california",
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Newsom vetoes Senate Bill 1047",
      "summary": "Veto: Governor Newsom returned Senate Bill 1047 unsigned, a bill on safeguards for developers of large artificial intelligence models and a Board of Frontier Models.",
      "says": null,
      "says_kind": "silent",
      "says_lang": null,
      "details": "The letter is dated September 29, 2024, and begins 'I am returning Senate Bill 1047 without my signature.' It says the bill would require developers of large artificial intelligence models, and those providing computing power to train them, to put safeguards in place, and would establish a Board of Frontier Models.",
      "license": null,
      "source_url": "https://www.gov.ca.gov/wp-content/uploads/2024/09/SB-1047-Veto-Message.pdf",
      "source_label": "Governor Newsom, SB 1047 veto message",
      "source_type": "agency",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-09-25-llama-3-2",
      "url": "https://maple.ai/open-weights#2024-09-25-llama-3-2",
      "date": "2024-09-25",
      "date_precision": "day",
      "actor": "Meta",
      "entity": "llama",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Llama 3.2 community license",
      "summary": "Release: Llama 3.2 weights under the Llama 3.2 Community License Agreement, which includes a 700 million monthly active user clause.",
      "says": "including machine-learning model code, trained model weights, inference-enabling code, training-enabling code, fine-tuning enabling code",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The license states “Llama 3.2 Version Release Date: September 25, 2024.” No clause banning use of outputs to improve another large language model was found in this file. The 700 million monthly active user clause remains. The acceptable-use policy linked from the license is https://www.llama.com/llama3_2/use-policy. The llama-models README table lists Llama 3.2 and Llama 3.2-Vision on the same date.",
      "license": {
        "name": "LLAMA 3.2 COMMUNITY LICENSE AGREEMENT",
        "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/LICENSE",
        "commercial": "If, on the Llama 3.2 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.",
        "previous": {
          "label": "LLAMA 3.1 COMMUNITY LICENSE AGREEMENT",
          "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/LICENSE",
      "source_label": "llama-models LICENSE, Llama 3.2",
      "source_type": "license",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2024-07-30-ntia-open-model-weights-report",
      "url": "https://maple.ai/open-weights#2024-07-30-ntia-open-model-weights-report",
      "date": "2024-07-30",
      "date_precision": "day",
      "actor": "National Telecommunications and Information Administration",
      "entity": null,
      "category": "report",
      "category_label": "Reports and plans",
      "title": "NTIA report on open weight models",
      "summary": "Report: NTIA states that current evidence is not sufficient to determine that restrictions on open weight models are warranted, or that they never will be.",
      "says": "current evidence is not sufficient to definitively determine either that restrictions on such open weight models are warranted, or that restrictions will never be appropriate in the future.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The report page is dated July 30, 2024, and says the report responds to section 4.6 of the October 2023 executive order after a February 2024 request for comment that drew 332 comments. It recommends monitoring and keeping open the possibility of later limits on model-weight availability.",
      "license": null,
      "source_url": "https://www.ntia.gov/issues/artificial-intelligence/open-model-weights-report",
      "source_label": "NTIA, Dual-Use Foundation Models with Widely Available Model Weights Report",
      "source_type": "agency",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2024-07-23-llama-3-1-license-change",
      "url": "https://maple.ai/open-weights#2024-07-23-llama-3-1-license-change",
      "date": "2024-07-23",
      "date_precision": "day",
      "actor": "Meta",
      "entity": "llama",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Llama 3.1 license change",
      "summary": "License change: Llama 3.1 weights for download; the license allows Llama outputs, including the 405B, to be used to improve other models.",
      "says": "We’ve also made changes to our license, allowing developers to use the outputs from Llama models—including the 405B—to improve other models.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The license states “Llama 3.1 Version Release Date: July 23, 2024.” The ban on using outputs to improve another large language model is not in this file. Section 1.b.i allows use of the materials or outputs “to create, train, fine tune, or otherwise improve an AI model” when the name starts with “Llama” and “Built with Llama” is displayed. The 700 million monthly active user clause remains. The post says the models are available for download on llama.meta.com and Hugging Face.",
      "license": {
        "name": "LLAMA 3.1 COMMUNITY LICENSE AGREEMENT",
        "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE",
        "commercial": "If, on the Llama 3.1 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.",
        "previous": {
          "label": "META LLAMA 3 COMMUNITY LICENSE AGREEMENT",
          "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://ai.meta.com/blog/meta-llama-3-1/",
      "source_label": "Meta AI blog, Introducing Llama 3.1",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-07-12-eu-ai-act-published",
      "url": "https://maple.ai/open-weights#2024-07-12-eu-ai-act-published",
      "date": "2024-07-12",
      "date_precision": "day",
      "actor": "European Parliament and Council",
      "entity": "eu-ai-act",
      "category": "law",
      "category_label": "Laws and bills",
      "title": "AI Act published in the Official Journal",
      "summary": "Law: Regulation (EU) 2024/1689 is published in the Official Journal, including rules on AI systems released under free and open-source licenses.",
      "says": "This Regulation does not apply to AI systems released under free and open-source licences, unless they are placed on the market or put into service as high-risk AI systems or as an AI system that falls under Article 5 or 50.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The act is dated June 13, 2024. The journal issue is dated July 12, 2024. Article 113 says it enters into force on the twentieth day after publication and applies Chapter V, on general-purpose AI models, from August 2, 2025, except Article 101. Article 111(3) gives providers of general-purpose AI models placed on the market before August 2, 2025 until August 2, 2027 to comply.",
      "license": null,
      "source_url": "https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32024R1689",
      "source_label": "Official Journal, Regulation (EU) 2024/1689",
      "source_type": "journal",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-06-07-qwen2-license-split",
      "url": "https://maple.ai/open-weights#2024-06-07-qwen2-license-split",
      "date": "2024-06-07",
      "date_precision": "day",
      "actor": "Qwen Team",
      "entity": "qwen",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Qwen2 license split",
      "summary": "License change: Qwen2-72B remains under the Qianwen License, and Qwen2-0.5B, 1.5B, 7B, and 57B-A14B move to Apache 2.0.",
      "says": "This time, we change the licenses of our models to different ones. While Qwen2-72B as well as its instruction-tuned models still uses the original Qianwen License, all other models, including Qwen2-0.5B, Qwen2-1.5B, Qwen2-7B, and Qwen2-57B-A14B, turn to adopt Apache 2.0!",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The license object above is the Apache 2.0 file on Qwen2-7B. It does not cover Qwen2-72B. The post says Qwen2-72B and its instruction-tuned models still use the original Qianwen License. That file’s 100 million monthly active user clause is the one quoted on the August 3, 2023 Tongyi Qianwen entry. The Qwen2-7B LICENSE appendix reads “Copyright 2024 Alibaba Cloud.”",
      "license": {
        "name": "Apache License, Version 2.0",
        "url": "https://huggingface.co/Qwen/Qwen2-7B/blob/main/LICENSE",
        "commercial": "each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form.",
        "previous": {
          "label": "Tongyi Qianwen LICENSE AGREEMENT",
          "url": "https://huggingface.co/Qwen/Qwen-7B/blob/main/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://qwenlm.github.io/blog/qwen2/",
      "source_label": "Qwen blog, Qwen2",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-05-29-mistral-mnpl",
      "url": "https://maple.ai/open-weights#2024-05-29-mistral-mnpl",
      "date": "2024-05-29",
      "date_precision": "day",
      "actor": "Mistral AI",
      "entity": "mistral",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Codestral non-production license",
      "summary": "Release: Codestral under the Mistral AI Non-Production License for non-production use, with Apache 2.0 releases continuing.",
      "says": "Codestral is released today under this new license",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The May 29, 2024 post says: “This license allows developers to use our technology for non-commercial purposes and to support research work.” It also says: “Note that we will keep releasing models and code under Apache 2.0 as we progressively consolidate two families of products released under Apache 2.0 and the MNPL.” The license limits use to “testing, research, Personal, or evaluation purposes in Non-Production Environments,” and its definition of the model includes “instructed checkpoints, parameters.”",
      "license": {
        "name": "Mistral AI Non-Production License",
        "url": "https://mistral.ai/licenses/MNPL-0.1.md",
        "commercial": "You shall not supply the Mistral Models or Derivatives in the course of a commercial activity, whether in return for payment or free of charge, in any medium or form, including but not limited to through a hosted or managed service (e.g. SaaS, cloud instances, etc.), or behind a software layer.",
        "previous": {
          "label": "Apache License, Version 2.0",
          "url": "https://www.apache.org/licenses/LICENSE-2.0.txt"
        },
        "weights": true
      },
      "source_url": "https://mistral.ai/news/mistral-ai-non-production-license-mnpl/",
      "source_label": "Mistral AI news, MNPL",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-04-18-llama-3",
      "url": "https://maple.ai/open-weights#2024-04-18-llama-3",
      "date": "2024-04-18",
      "date_precision": "day",
      "actor": "Meta",
      "entity": "llama",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Llama 3 weights released",
      "summary": "Release: Llama 3 weights under the Meta Llama 3 Community License Agreement, including a 700 million monthly active user clause.",
      "says": "Today, we’re introducing Meta Llama 3, the next generation of our state-of-the-art open source large language model.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The license states “Meta Llama 3 Version Release Date: April 18, 2024.” It keeps the clause barring use of the materials or their output “to improve any other large language model (excluding Meta Llama 3 or derivative works thereof).” The acceptable-use policy linked from the license is https://llama.meta.com/llama3/use-policy. Attribution in the license uses “Built with Meta Llama 3.”",
      "license": {
        "name": "META LLAMA 3 COMMUNITY LICENSE AGREEMENT",
        "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3/LICENSE",
        "commercial": "If, on the Meta Llama 3 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.",
        "previous": {
          "label": "LLAMA 2 COMMUNITY LICENSE AGREEMENT",
          "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama2/LICENSE"
        },
        "weights": true
      },
      "source_url": "https://ai.meta.com/blog/meta-llama-3/",
      "source_label": "Meta AI blog, Meta Llama 3",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-02-26-ntia-widely-available-weights-rfc",
      "url": "https://maple.ai/open-weights#2024-02-26-ntia-widely-available-weights-rfc",
      "date": "2024-02-26",
      "date_precision": "day",
      "actor": "National Telecommunications and Information Administration",
      "entity": null,
      "category": "report",
      "category_label": "Reports and plans",
      "title": "NTIA requests comment on model weights",
      "summary": "Request for comment: NTIA asks how to define widely available model weights, which the notice calls open foundation models, and what rules might apply.",
      "says": "Dual use foundation models with widely available weights (referred to here as open foundation models) could play a key role in fostering growth among less resourced actors, helping to widely share access to AI's benefits.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The notice, Docket No. 240216-0052, RIN 0660-XC060, implements section 4.6 of the October 30, 2023 executive order. Comments were due March 27, 2024. The notice also says widely available weights may pose risks, including affirmative misuse and lack of monitoring.",
      "license": null,
      "source_url": "https://www.govinfo.gov/content/pkg/FR-2024-02-26/html/2024-03763.htm",
      "source_label": "Federal Register, 89 FR 14059",
      "source_type": "register",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2024-02-21-gemma",
      "url": "https://maple.ai/open-weights#2024-02-21-gemma",
      "date": "2024-02-21",
      "date_precision": "day",
      "actor": "Google",
      "entity": "gemma",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Gemma 2B and 7B weights",
      "summary": "Release: Gemma 2B and Gemma 7B model weights, with terms the post says permit commercial usage and distribution for all organizations.",
      "says": "Terms of use permit responsible commercial usage and distribution for all organizations, regardless of size.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The post says: “We’re releasing model weights in two sizes: Gemma 2B and Gemma 7B.” It calls them “open models.” The terms page incorporates the Gemma Prohibited Use Policy at ai.google.dev/gemma/prohibited_use_policy. Distribution is to carry the notice “Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms.”",
      "license": {
        "name": "Gemma Terms of Use",
        "url": "https://ai.google.dev/gemma/terms",
        "commercial": "You may use, reproduce, modify, Distribute, perform or display any of the Gemma Services only in accordance with the terms of this Agreement",
        "previous": null,
        "weights": true
      },
      "source_url": "https://blog.google/innovation-and-ai/technology/developers-tools/gemma-open-models/",
      "source_label": "Google blog, Gemma open models",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2023-10-30-eo-14110-widely-available-weights",
      "url": "https://maple.ai/open-weights#2023-10-30-eo-14110-widely-available-weights",
      "date": "2023-10-30",
      "date_precision": "day",
      "actor": "The President",
      "entity": null,
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Order to study widely available model weights",
      "summary": "Order: Section 4.6 directs Commerce, through NTIA, to seek comment and report on dual-use foundation models whose model weights are widely available.",
      "says": "When the weights for a dual-use foundation model are widely available--such as when they are publicly posted on the internet--there can be substantial benefits to innovation, but also substantial security risks, such as the removal of safeguards within the model.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "Executive Order 14110 is dated October 30, 2023, and was published in the Federal Register on November 1, 2023. Section 4.6 gives the Secretary 270 days to solicit public input and submit a report to the President.",
      "license": null,
      "source_url": "https://www.govinfo.gov/content/pkg/FR-2023-11-01/html/2023-24283.htm",
      "source_label": "Federal Register, Executive Order 14110, 88 FR 75191",
      "source_type": "register",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2023-10-25-uk-open-source-generative-ai",
      "url": "https://maple.ai/open-weights#2023-10-25-uk-open-source-generative-ai",
      "date": "2023-10-25",
      "date_precision": "day",
      "actor": "Department for Science, Innovation and Technology",
      "entity": null,
      "category": "safety",
      "category_label": "Safety findings",
      "title": "Paper on open-source generative AI",
      "summary": "Report: A DSIT discussion paper for the 2023 AI Safety Summit describes open-source generative AI and access to model development.",
      "says": "Open-source generative AI is facilitating rapid proliferation and increasing democratisation of generative AI by reducing the barriers to entry for developing models.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "Paragraph 4 says open-source models had mostly lagged frontier models and were highly unlikely to be more capable than leading commercial frontier models by 2025, and that proliferation brings safety and security implications where safeguards are not effective. The parent page says the paper does not represent government policy.",
      "license": null,
      "source_url": "https://assets.publishing.service.gov.uk/media/653932db80884d0013f71b15/generative-ai-safety-security-risks-2025-annex-b.pdf",
      "source_label": "GOV.UK, generative AI risks to 2025, Annex B",
      "source_type": "agency",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2023-09-27-mistral-7b",
      "url": "https://maple.ai/open-weights#2023-09-27-mistral-7b",
      "date": "2023-09-27",
      "date_precision": "day",
      "actor": "Mistral AI",
      "entity": "mistral",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Mistral 7B under Apache 2.0",
      "summary": "Release: Mistral 7B under the Apache 2.0 license, which the post says can be used without restrictions.",
      "says": "We're releasing Mistral 7B under the Apache 2.0 license, it can be used without restrictions.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The September 27, 2023 post is the release notice and includes a download link. The Hugging Face repo mistralai/Mistral-7B-v0.1 lists license apache-2.0 and weight files, and it has no LICENSE file among its siblings.",
      "license": {
        "name": "Apache License, Version 2.0",
        "url": "https://www.apache.org/licenses/LICENSE-2.0.txt",
        "commercial": "each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form.",
        "previous": null,
        "weights": true
      },
      "source_url": "https://mistral.ai/news/announcing-mistral-7b",
      "source_label": "Mistral AI news, Announcing Mistral 7B",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2023-08-03-qwen-tongyi-license",
      "url": "https://maple.ai/open-weights#2023-08-03-qwen-tongyi-license",
      "date": "2023-08-03",
      "date_precision": "day",
      "actor": "Alibaba Cloud",
      "entity": "qwen",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Tongyi Qianwen license agreement",
      "summary": "Release: Tongyi Qianwen materials, including trained model weights, under a license with a 100 million monthly active user clause.",
      "says": "consisting of trained model weights, parameters (including optimizer states), machine-learning model code, inference-enabling code, training-enabling code, fine-tuning enabling code",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The file states “Tongyi Qianwen Release Date: August 3, 2023” and names Alibaba Cloud. It also says: “You can not use the Materials or any output therefrom to improve any other large language model (excluding Tongyi Qianwen or derivative works thereof).” Materials include “trained model weights.”",
      "license": {
        "name": "Tongyi Qianwen LICENSE AGREEMENT",
        "url": "https://huggingface.co/Qwen/Qwen-7B/blob/main/LICENSE",
        "commercial": "If you are commercially using the Materials, and your product or service has more than 100 million monthly active users, You shall request a license from Us. You cannot exercise your rights under this Agreement without our express authorization.",
        "previous": null,
        "weights": true
      },
      "source_url": "https://huggingface.co/Qwen/Qwen-7B/blob/main/LICENSE",
      "source_label": "Qwen-7B LICENSE, Tongyi Qianwen LICENSE AGREEMENT",
      "source_type": "license",
      "verified_on": "2026-10-20",
      "corrections": []
    },
    {
      "id": "2023-07-18-llama-2",
      "url": "https://maple.ai/open-weights#2023-07-18-llama-2",
      "date": "2023-07-18",
      "date_precision": "day",
      "actor": "Meta",
      "entity": "llama",
      "category": "release",
      "category_label": "Releases and licenses",
      "title": "Llama 2 weights released",
      "summary": "Release: Llama 2 weights for research and commercial use under the Llama 2 Community License Agreement, which includes a 700 million monthly active user clause.",
      "says": "We’re now ready to open source the next version of Llama 2 and are making it available free of charge for research and commercial use. We’re including model weights and starting code for the pretrained model and conversational fine-tuned versions too.",
      "says_kind": "quote",
      "says_lang": null,
      "details": "The license is headed “Llama 2 Version Release Date: July 18, 2023.” It also says: “You will not use the Llama Materials or any output or results of the Llama Materials to improve any other large language model (excluding Llama 2 or derivative works thereof).” An Acceptable Use Policy is at https://ai.meta.com/llama/use-policy.",
      "license": {
        "name": "LLAMA 2 COMMUNITY LICENSE AGREEMENT",
        "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama2/LICENSE",
        "commercial": "If, on the Llama 2 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee's affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.",
        "previous": null,
        "weights": true
      },
      "source_url": "https://ai.meta.com/blog/llama-2/",
      "source_label": "Meta AI blog, Llama 2",
      "source_type": "company",
      "verified_on": "2026-10-16",
      "corrections": []
    },
    {
      "id": "2023-07-13-china-generative-ai-measures",
      "url": "https://maple.ai/open-weights#2023-07-13-china-generative-ai-measures",
      "date": "2023-07-13",
      "date_precision": "day",
      "actor": "Cyberspace Administration of China",
      "entity": null,
      "category": "law",
      "category_label": "Laws and bills",
      "title": "Interim rules for generative AI services",
      "summary": "Rule: Seven Chinese agencies publish interim measures for generative AI services offered to the public in China, effective August 15, 2023.",
      "says": null,
      "says_kind": "silent",
      "says_lang": "zh",
      "details": "The order, signed July 10, 2023 and posted July 13, 2023, says the measures apply to services that use generative AI to provide text, images, audio, or video to the public in China. Article 17 requires a security assessment and algorithm filing for services with public-opinion attributes or social-mobilization capacity. Article 22 defines a provider to include offering the service through a programmable interface.",
      "license": null,
      "source_url": "https://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm",
      "source_label": "CAC, Interim Measures for Generative AI Services",
      "source_type": "agency",
      "verified_on": "2026-10-20",
      "corrections": []
    }
  ]
}