Many observers regard “Llama,” a series of AI models developed by Meta Platforms, Inc., as both highly performant and cost-effective, with relatively permissive distribution conditions. As a result, its adoption for various systems and the development of derivative models appear to be expanding. However, because Meta’s CEO has publicly touted Llama as if it were “Open Source,” there has been an increasing misconception that Llama truly is Open Source. Moreover, despite the fact that the Llama License contains several problematic provisions, its ongoing adoption in a wide array of services may well give rise to serious legal complications. Consequently, this article addresses, first and foremost, whether the Llama License complies with Open Source standards. (Potential risks associated with actual model usage under the Llama License will be covered in a separate piece.)

Original Japanese Article: https://shujisado.com/2025/01/15/llama_is_not_opensource/


Conformance of Llama to Open Source

The Llama Community License Agreement (Llama License), which applies to Llama models, may initially appear to some—particularly those without extensive licensing or copyright knowledge—to be not so different from an Open Source license. Indeed, while it imposes various restrictions (e.g., on usage, reproduction, distribution, and copying), the Llama License does afford a certain degree of freedom and expressly permits creating derivative works. In that sense, it shares some characteristics commonly associated with Open Source licenses.

Yet, comparing the Llama License to the Open Source Definition (OSD) promulgated by the Open Source Initiative (the only organization that certifies licenses as truly Open Source) quickly makes clear that the Llama License lacks compliance with Open Source standards in numerous respects. Below, we examine the relevant provisions of the Llama License to highlight specific reasons why Llama cannot be considered Open Source.

1. Principal Reasons Denying Open Source Status

There are multiple grounds for concluding that the Llama License is not Open Source. However, two fundamental issues reflecting the agreement’s goals stand out: (1) restrictions on usage by major corporations, and (2) the presence of an Acceptable Use Policy integrated into the contract.

1.1. The 700 Million Monthly Active Users Restriction

Llama license, Section 2 states:
“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 …”

Section 2 of the Llama License straightforwardly decrees that any entity with 700 million monthly active users cannot employ the Llama model without Meta’s express approval. In effect, it appears aimed at preventing usage of Llama by large competitors that could threaten Meta’s market position. However, such a restriction contradicts OSD 5, which prohibits “discrimination against persons or groups.”

Some might contend that well-funded technology giants are already profitable enough for this limitation to be immaterial, but it must be remembered that Open Source is grounded in affording freedom to all parties regardless of organizational size or business type. Furthermore, from a commercial perspective, the threshold might inhibit smaller developers incorporating Llama-based services or derivative works from selling to or merging with major corporations, a potential business concern.

Note as well that some users misread the restriction as applying only to user counts in a specific product or service using Llama. In fact, it refers to the entire corporate group’s user base: if multiple services exist, their user numbers must be aggregated. Likewise, the user totals of any parent or subsidiary exceeding 50% ownership also factor in. (This issue is treated more fully in a separate discussion.)

1.2. The Incorporated Acceptable Use Policy

Section 1.b.iv provides:

“Your use of the Llama Materials must … adhere to the Acceptable Use Policy for the Llama Materials (available at https://llama.meta.com/llama3_1/use-policy), which is hereby incorporated by reference into this Agreement.”

Meta publishes, in a separate document, an Acceptable Use Policy that Llama model users must follow when employing Llama. Section 1.b.iv specifies that this AUP is “incorporated by reference” into the Llama License, effectively making the policy part of the Llama License itself.

Broadly, the AUP prohibits (1) uses that are illegal or infringe third-party rights, (2) developing or facilitating activities that might cause death or bodily harm, (3) willful deception of others, and (4) failing to disclose known AI-system risks to end users. Although these clauses target socially undesirable behavior, by restricting also lawful activities under the guise of “ethical” policy measures, the AUP runs afoul of OSD 6, which bars discrimination based on “fields of endeavor.”

One might ask, “Why is it problematic to ensure that individuals employ AI models correctly?” Yet what exactly constitutes correct usage, and who decides it? Ultimately, Meta, at its own discretion, interprets the numerous “undesirable actions” vaguely outlined in the AUP. Furthermore, Section 7 of the Llama License specifies California law as the governing law, and Meta is a U.S. company; naturally, ethical standards would thus reflect American culture and practice. While this may not heavily impact organizations or individuals within the United States, it can result in unintentional violations by licensees from countries like Japan, whose legal and cultural norms differ. A scenario reminiscent of Visa’s 2024 suspension of certain card payment services could plausibly arise. Additionally, Section 6 stipulates that Meta may terminate the contract at any time upon concluding the licensee is in breach, with no grace period akin to that found in the GNU GPL.

1.3. AUP Subject to Ongoing Expansion

The AUP-related concerns in the preceding section are compounded by another issue. The policy is “incorporated by reference,” as stated in the contract text, meaning that the policy displayed at the specified URL is perpetually valid and may be unilaterally updated by Meta at any time. Since the AUP forms part of the Llama License, even the license’s core provisions can effectively shift if Meta modifies the policy—implicating OSD 7, on the “distribution” of the license.

Meta could, for instance, retroactively insert clauses forcing many companies to negotiate separate paid agreements, provided that such amendments are conceivably justified under principles of good faith or fair dealing. In effect, this arrangement predates whether it even satisfies or contravenes the OSD’s narrower strictures—it’s already structurally problematic.

Some might insist that Meta, as a publicly traded enterprise, would not undertake such extreme conduct. Yet with Llama 3.2, the Acceptable Use Policy added a provision under Section 1.a that revokes from EU users the rights “to use, reproduce, distribute, copy, create derivative works, and modify the Llama Materials.” This arguably goes well beyond any standard “Open Source” debate. Moreover, embedding such a clause—one that negates a core license grant—in the Acceptable Use Policy, rather than the license text proper, raises additional questions about the validity of the entire mechanism.

“With respect to any multimodal models included in Llama 3.2, the rights granted under Section 1(a) of the Llama 3.2 Community License Agreement are not being granted to you if you are an individual domiciled in, or a company with a principal place of business in, the European Union.”

Llama 3.2’s AUP also adds other clauses, ostensibly concerning safety, that might inhibit meaningful examination of the model’s behavior. These constraints may well curtail a researcher’s freedom to study how Llama operates, a principle at the core of Open Source. For instance:

h. Engage in any action, or facilitate any action, to intentionally circumvent or remove usage restrictions or other safety measures, or to enable functionality disabled by Meta

5. Interact with third party tools, models, or software designed to generate unlawful content or engage in unlawful or harmful conduct and/or represent that the outputs of such tools, models, or software are associated with Meta or Llama 3.2

Each of these provisions, in effect, could hamper evaluating the safety of the Llama model or reduce transparency in its research. Consequently, they may well violate fundamental Open Source norms.

2. Additional Reasons Denying Open Source Conformity

The reasons outlined above already provide ample grounds to conclude that Llama is not Open Source. However, the Llama License contains other provisions that likewise conflict with typical Open Source licensing requirements. Below is a list and explanation of those additional issues.

2.1. Branding/Naming Restrictions on Derivative Works

Section 1.b.i:

“Prominently display ‘Built with Llama.’”

“If you use the Llama Materials or any outputs or results of the Llama Materials to create, train, fine tune, or otherwise improve an AI model, which is distributed or made available, you shall also include ‘Llama’ at the beginning of any such AI model name.”

Under Section 1.b.i, in addition to requiring that the Llama License text accompany any distributed derivative work, the developer must conspicuously display “Built with Llama” on relevant product pages and documentation, and prepend the word “Llama” to the name of any derivative model. This likely violates OSD 8, which forbids licenses that are restricted to certain products only. Generally speaking, an Open Source license allows for correct attribution of copyrights but does not require specific naming or branding. By mandating distinctive brand usage, the Llama License imposes constraints on the creation and distribution of derivatives.

One might regard a mere naming requirement as trivial. However, such an obligation persists as long as the contractual relationship lasts, irrespective of how extensive any derivative modifications might be. From Meta’s perspective, it confers the added advantage of derivative-model developers enhancing Meta’s brand visibility.

2.2. Prohibition on Assignment of the Contract

Section 1.a outlines the rights granted to the licensee, clarifying that the Llama License is “non-transferable.” This likely contravenes OSD 1—the principle of free redistribution.

A question may arise: “What does ‘non-transferable’ mean if one can create derivative models and even distribute them?” Note that when a party distributes a derivative model under the Llama License, the recipient user enters into a new Llama License agreement with Meta, thereby not receiving a transferred contract from the developer. Consequently, that scenario is not the problem. Instead, issues arise where a licensee, having earlier entered into the Llama License, attempts to transfer its contractual status to a third party—for example, in mergers, acquisitions, or sales of businesses—particularly if the original licensee entity ceases to exist. Unless Meta consents, the Llama License deems such a transfer impermissible, creating complications for normal corporate transactions.

2.3. Contract Extending Even to Model Outputs

Again referring to Section 1.b.i:

“If you use the Llama Materials or any outputs or results of the Llama Materials to create, train, fine tune, or otherwise improve an AI model, which is distributed or made available, you shall also include ‘Llama’ at the beginning of any such AI model name.”

This clause, newly added in version 3.1 of the Llama License, previously highlighted in these discussions, restricts the naming not only of derivatives directly built from the Llama Materials but also of those built using Llama’s outputs or results as training data. Put differently, it enforces license conditions extending into any models trained in part on Llama’s generative outputs.

Some might analogize this to “copyleft” obligations under the GNU GPL, etc. However, actual copyleft rests on recognized copyright coverage, whereas the notion that the model’s output could be subject to the model’s own intellectual property rights remains questionable. While there is some argument for derivative liability in training data usage, it remains that the model’s outputs are probabilistic in nature. Attempting to impose license terms on such statistically generated content could be considered novel, if not overreaching. Traditionally, Open Source licenses rely on the global copyright system to confer freedoms; by contrast, the Llama License seeks to regulate domains extending beyond pure copyright. It thereby runs contrary to the freedom that Open Source aims to advance.

2.4. Additional Trademark Restrictions

Section 5.a clarifies that no trademark rights are granted. Many Open Source licenses likewise disclaim trademark rights, so this in and of itself is not necessarily a problem. Nevertheless, as noted, Llama insists that any derivative model incorporate “Llama” in its name, and Section 5.a further imposes Meta’s brand guidelines on such limited trademark use—granting all goodwill arising from that usage to Meta. Although this may be somewhat tangential to the immediate question of Open Source compliance, it exemplifies how restrictive the Llama License can be.

2.5. The License’s Essential Character as a Contract

An Open Source license is, at bottom, a unilateral permission for the exercise of copyright—one that has historically enabled communities to expand ecosystems via a simple grant of rights. While certain lawsuits have recognized a contractual aspect to open licenses, in day-to-day practice, they are widely regarded as unilateral grants by the rights holder.

Conversely, the Llama License is styled from the outset as a contract: The user’s acceptance (e.g., via a click-through) and subsequent usage of Llama reflect mutual consideration and an exchange of obligations that go beyond simple copyright permission. In effect, the arrangement rests on U.S. contract law between Meta and each licensee.

Although this approach does not necessarily disqualify an arrangement from being Open Source, with Llama, it amplifies the binding obligations outlined earlier in ways that may exceed a literal reading of the text, reinforcing contractual force that could narrow the freedoms championed by Open Source.

3. Open Source AI

Taking these points in conjunction—the user-threshold restriction, unilateral Acceptable Use Policy, brand and naming constraints, intellectual property claims extending to outputs, etc.—it becomes evident that the Llama License violates the majority of the OSD’s clauses, and that Llama is not Open Source. However, some argue that AI models differ from conventional software and need not adhere to OSI’s definition of Open Source. They typically cite OSD 2 (“Source Code”) to suggest that the OSI’s Open Source criteria only apply to software with accessible source code.

Partly to address such claims, volunteer experts worldwide (including the present author) collaborated over two years to publish the Open Source AI Definition (OSAID) by October 2024. OSAID sets the conditions under which an AI system—comprising a model, code, and data—may call itself “Open Source AI.” It does not assess Open Source status solely for an isolated model component. Yet the recurring question, “What counts as source code in an AI model?” was debated frequently. Across numerous jurisdictions, the trained model—being merely a final product of computations—may not retain any copyright from the original training data, especially if the data does not actually “control” the model. Nonetheless, there is consensus that fully revealing the model’s code, weights, and training data’s information aligns best with open principles and remains the aspiration for AI systems that would claim to be Open Source. That is effectively the stance of OSAID as well.

It is thus plain that the Llama License does not meet the established “Open Source Definition.” Even if, hypothetically, the Llama License did, the Llama model’s code and data’s information are not fully disclosed; hence, Llama would still not qualify as an Open Source AI system.

4. Conclusion

Both the Llama model and the Llama License fall short of Open Source principles and the definition of Open Source. Meta, of course, may formulate its licenses as it sees fit; however, marketing a non–Open Source license as if it were Open Source undermines the value of the Open Source community and ecosystem, which must not be allowed.

Likewise, anyone planning to adopt Llama must clearly understand its licensing nature and consider the impact on end users of Llama-based services or derivative models. In particular, the powerful, contract-based Acceptable Use Policy can have more extensive effects than some realize, potentially rendering significant investments futile if overlooked.

Finally, the additional contractual intricacies of the Llama License that go beyond Llama’s Open Source status are, as noted, addressed in a separate article.


LLAMA 3.2 COMMUNITY LICENSE AGREEMENT

Llama 3.2 Version Release Date: September 25, 2024

Agreement” means the terms and conditions for use, reproduction, distribution and modification of the Llama Materials set forth herein.

Documentation” means the specifications, manuals and documentation accompanying Llama 3.2 distributed by Meta at https://llama.com/docs/overview.

Licensee” or “you” means you, or your employer or any other person or entity (if you are entering into this Agreement on such person or entity’s behalf), of the age required under applicable laws, rules or regulations to provide legal consent and that has legal authority to bind your employer or such other person or entity if you are entering in this Agreement on their behalf.

Llama 3.2” means the foundational large language models and software and algorithms, including machine-learning model code, trained model weights, inference-enabling code, training-enabling code, fine-tuning enabling code and other elements of the foregoing distributed by Meta at https://llama.com/llama-downloads.

Llama Materials” means, collectively, Meta’s proprietary Llama 3.2 and Documentation (and any portion thereof) made available under this Agreement.

Meta” or “we” means Meta Platforms Ireland Limited (if you are located in or, if you are an entity, your principal place of business is in the EEA or Switzerland) and Meta Platforms, Inc. (if you are located outside of the EEA or Switzerland).

By clicking “I Accept” below or by using or distributing any portion or element of the Llama Materials, you agree to be bound by this Agreement.

1. License Rights and Redistribution.

a. Grant of Rights. You are granted a non-exclusive, worldwide, non-transferable and royalty-free limited license under Meta’s intellectual property or other rights owned by Meta embodied in the Llama Materials to use, reproduce, distribute, copy, create derivative works of, and make modifications to the Llama Materials.

b. Redistribution and Use.

i. If you distribute or make available the Llama Materials (or any derivative works thereof), or a product or service (including another AI model) that contains any of them, you shall (A) provide a copy of this Agreement with any such Llama Materials; and (B) prominently display “Built with Llama” on a related website, user interface, blogpost, about page, or product documentation. If you use the Llama Materials or any outputs or results of the Llama Materials to create, train, fine tune, or otherwise improve an AI model, which is distributed or made available, you shall also include “Llama” at the beginning of any such AI model name.

ii. If you receive Llama Materials, or any derivative works thereof, from a Licensee as part of an integrated end user product, then Section 2 of this Agreement will not apply to you.

iii. You must retain in all copies of the Llama Materials that you distribute the following attribution notice within a “Notice” text file distributed as a part of such copies: “Llama 3.2 is licensed under the Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.”

iv. Your use of the Llama Materials must comply with applicable laws and regulations (including trade compliance laws and regulations) and adhere to the Acceptable Use Policy for the Llama Materials (available at https://llama.com/llama3_2/use-policy), which is hereby incorporated by reference into this Agreement.

2. Additional Commercial Terms. 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.

3. Disclaimer of Warranty. UNLESS REQUIRED BY APPLICABLE LAW, THE LLAMA MATERIALS AND ANY OUTPUT AND RESULTS THEREFROM ARE PROVIDED ON AN “AS IS” BASIS, WITHOUT WARRANTIES OF ANY KIND, AND META DISCLAIMS ALL WARRANTIES OF ANY KIND, BOTH EXPRESS AND IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. YOU ARE SOLELY RESPONSIBLE FOR DETERMINING THE APPROPRIATENESS OF USING OR REDISTRIBUTING THE LLAMA MATERIALS AND ASSUME ANY RISKS ASSOCIATED WITH YOUR USE OF THE LLAMA MATERIALS AND ANY OUTPUT AND RESULTS.

4. Limitation of Liability. IN NO EVENT WILL META OR ITS AFFILIATES BE LIABLE UNDER ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, TORT, NEGLIGENCE, PRODUCTS LIABILITY, OR OTHERWISE, ARISING OUT OF THIS AGREEMENT, FOR ANY LOST PROFITS OR ANY INDIRECT, SPECIAL, CONSEQUENTIAL, INCIDENTAL, EXEMPLARY OR PUNITIVE DAMAGES, EVEN IF META OR ITS AFFILIATES HAVE BEEN ADVISED OF THE POSSIBILITY OF ANY OF THE FOREGOING.

5. Intellectual Property.

a. No trademark licenses are granted under this Agreement, and in connection with the Llama Materials, neither Meta nor Licensee may use any name or mark owned by or associated with the other or any of its affiliates, except as required for reasonable and customary use in describing and redistributing the Llama Materials or as set forth in this Section 5(a). Meta hereby grants you a license to use “Llama” (the “Mark”) solely as required to comply with the last sentence of Section 1.b.i. You will comply with Meta’s brand guidelines (currently accessible at https://about.meta.com/brand/resources/meta/company-brand/). All goodwill arising out of your use of the Mark will inure to the benefit of Meta.

b. Subject to Meta’s ownership of Llama Materials and derivatives made by or for Meta, with respect to any derivative works and modifications of the Llama Materials that are made by you, as between you and Meta, you are and will be the owner of such derivative works and modifications.

c. If you institute litigation or other proceedings against Meta or any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Llama Materials or Llama 3.2 outputs or results, or any portion of any of the foregoing, constitutes infringement of intellectual property or other rights owned or licensable by you, then any licenses granted to you under this Agreement shall terminate as of the date such litigation or claim is filed or instituted. You will indemnify and hold harmless Meta from and against any claim by any third party arising out of or related to your use or distribution of the Llama Materials.

6. Term and Termination. The term of this Agreement will commence upon your acceptance of this Agreement or access to the Llama Materials and will continue in full force and effect until terminated in accordance with the terms and conditions herein. Meta may terminate this Agreement if you are in breach of any term or condition of this Agreement. Upon termination of this Agreement, you shall delete and cease use of the Llama Materials. Sections 3, 4 and 7 shall survive the termination of this Agreement.

7. Governing Law and Jurisdiction. This Agreement will be governed and construed under the laws of the State of California without regard to choice of law principles, and the UN Convention on Contracts for the International Sale of Goods does not apply to this Agreement. The courts of California shall have exclusive jurisdiction of any dispute arising out of this Agreement.

The Hidden Risks of NVIDIA’s Open Model License

Recently, regarding the open-weights AI model “Nemotron 3” released by NVIDIA, there are scattered media reports mistakenly describing it as open source. Because there is concern that these reports encourage ignoring the usage risks of the NVIDIA Open Model License Agreement (version dated October 24, 2025; hereinafter referred to as the NVIDIA License), which is…

The Current State of the Theory that GPL Propagates to AI Models Trained on GPL Code

When GitHub Copilot was launched in 2021, the fact that its training data included a vast amount of Open Source code publicly available on GitHub attracted significant attention, sparking lively debates regarding licensing. While there were issues concerning conditions such as attribution required by most licenses, there was a particularly high volume of discourse suggesting…

The Legal Hack: Why U.S. Law Sees Open Source as “Permission,” Not a Contract

In Japan, the common view is to treat an Open Source license as a license agreement, or a contract. This is also the case in the EU. However, in the United States—the origin point for almost every aspect of Open Source—an Open Source license has long been considered not a contract, but a “unilateral permission”…

Evaluating OpenMDW: A Revolution for Open AI, or a License to Openwash?

Although the number of AI models distributed under Open Source licenses is increasing, it can be said that AI systems in which all related components, including training data, are open are still in a developmental stage, even as a few promising systems have emerged. In this context, this past May, the Linux Foundation, in collaboration…

A Curious Phenomenon with Gemma Model Outputs and License Propagation

While examining the licensing details of Google’s Gemma model, I noticed a potentially puzzling phenomenon: you can freely assign a license to the model’s outputs, yet depending on how those outputs are used, the original Terms of Use might suddenly propagate to the resulting work. Outputs vs. Model Derivatives The Gemma Terms of Use distinguish…

Should ‘Open Source AI’ Mean Exposing All Training Data?

DeepSeek has had a major global impact. This appears to stem not only from the emergence of a new force in China that threatens the dominance of major U.S. AI vendors, but also from the fact that the AI model itself is being distributed under the MIT License, which is an Open Source license. Nevertheless,…

Significant Risks in Using AI Models Governed by the Llama License

Although it has already been explained that the Llama model and the Llama License (Llama Community License Agreement) do not, in any sense, qualify as Open Source, it bears noting that the Llama License contains several additional issues. While not directly relevant to whether it meets Open Source criteria, these provisions may nonetheless cause the…

The Hidden Traps in Meta’s Llama License

— An Explanation of Llama’s Supposed “Open Source” Status and the Serious Risks of Using Models under the Llama License — It is widely recognized—despite Meta’s CEO persistently promoting the notion that “Llama is Open Source”—that the Llama License is in fact not Open Source. Yet few individuals have clearly articulated the precise reasons why…