<!DOCTYPE html>
<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>The definition of "Derivative Materials" includes Output in those
      cases where the Output is for the purpose of creating of another
      model. </p>
    <p>We seem to be in agreement about the problem, an over-expansive
      definition of derivative works, but we have identified it for
      different reasons. I look at it theoretically and believe that the
      attempt to control Output at all is not justified because it's not
      limited to derivative works. It's particularly problematic for a
      model license, because the output is highly unlikely to be
      protected by copyright. You view it in operation and spot the
      problem by realizing that the definition will reach agents, which
      wasn't your intention. I think we get to the same place, just
      through different lenses.</p>
    <p>The unintended expansion is why tying a license scope to the
      exclusive rights of authors, rather than trying to define the
      scope by the type or purpose of a change, is, in my view, the
      sounder approach. </p>
    <p>If the attempt to control output that isn't a derivative work was
      removed, I believe I would probably agree the license is
      acceptable, although I'd have to review it again to be sure.</p>
    <p>Pam</p>
    <br>
    <div class="moz-signature">Pamela S. Chestek<br>
      Chestek Legal<br>
      4641 Post St. <br>
      Unit 4316<br>
      El Dorado Hills, CA 95762<br>
      +1 919-800-8033<br>
      pamela@chesteklegal<br>
      <a class="moz-txt-link-abbreviated" href="http://www.chesteklegal.com">www.chesteklegal.com</a><br>
      <br>
      <br>
    </div>
    <div class="moz-cite-prefix">On 12/5/2025 6:23 PM, Moming Duan
      wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:541EAF35-5686-48A5-8839-5808AA5E20FC@gmail.com">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      Hi Pam,
      <div><br>
      </div>
      <div><br>
      </div>
      <div>For clarification, the definition of Derivative Materials
        does not include the Output of the model (at least that is my
        intention). It includes derivative models created by
        transferring patterns from the Output. However, I am
        reconsidering this definition: as more AI agent systems are
        built on top of model outputs, I am concerned that including
        \u201cderivative models developed by transferring patterns of Output\u201d
        may unintentionally broaden the scope of what counts as a
        derivative, thereby restricting downstream innovation and
        conflicting with the Open Source spirit, which is not the
        intention of this license. </div>
      <div><br>
      </div>
      <div>I would appreciate further discussion on this issue,
        including the potential pros and cons of removing this
        Output-related element from the current definition. Thanks.</div>
      <div><br>
      </div>
      <div><br>
      </div>
      <div>Best,</div>
      <div>Moming<br id="lineBreakAtBeginningOfMessage">
        <div><br>
          <blockquote type="cite">
            <div>On Dec 6, 2025, at 01:22, Pamela Chestek
              <a class="moz-txt-link-rfc2396E" href="mailto:pamela@chesteklegal.com">&lt;pamela@chesteklegal.com&gt;</a> wrote:</div>
            <br class="Apple-interchange-newline">
            <div>
              <meta http-equiv="Content-Type"
                content="text/html; charset=UTF-8">
              <div> I renew my objections to this license and the
                Attribution-ShareAlike version for the same reasons I
                objected to the previous version, starting with this
                email,
                <a class="moz-txt-link-freetext"
href="https://lists.opensource.org/pipermail/license-review_lists.opensource.org/2025-May/005766.html"
                  moz-do-not-send="true">https://lists.opensource.org/pipermail/license-review_lists.opensource.org/2025-May/005766.html</a>,
                and as explained further in later emails in the thread.
                I do not believe that putting conditions on output of a
                model is workable and, where the output is not a
                derivative work under copyright law, it violates OSD9,
                "License Must Not Restrict Other Software." <br>
                <br>
                The change "Narrows the definition of 'Derivative
                Materials' by including the phrase: 'in order to
                replicate, approximate, or otherwise achieve functional
                behavior that is similar to the Model'" does not address
                this problem and, in fact, exacerbates it. Output that
                will "replicate, approximate, or otherwise achieve
                functional behavior that is similar to the Model"
                identifies output that is highly likely to not be a
                derivative work under any stretch of the imagination and
                therefore is well beyond the acceptable reach for an
                open source license.<br>
                <br>
                I do not believe these two versions of the license can
                be approved.<br>
                <br>
                Pam<br>
                <br>
                <div class="moz-signature">Pamela S. Chestek<br>
                  Chestek Legal<br>
                  4641 Post St.<br>
                  Unit 4316<br>
                  El Dorado Hills, CA 95762<br>
                  +1 919-800-8033<br>
                  <a
class="moz-txt-link-abbreviated moz-txt-link-freetext"
                    href="mailto:pamela@chesteklegal.com"
                    moz-do-not-send="true">pamela@chesteklegal.com</a><br>
                  <a class="moz-txt-link-abbreviated"
                    href="http://www.chesteklegal.com/"
                    moz-do-not-send="true">www.chesteklegal.com</a><br>
                  <br>
                  <a
href="https://calendly.com/pamela-chesteklegal/30min"
                    moz-do-not-send="true">Set a meeting with me</a></div>
                <div class="moz-cite-prefix">On 12/5/2025 9:04 AM, McCoy
                  Smith wrote:<br>
                </div>
                <blockquote type="cite"
cite="mid:e325347c-0c66-4294-bfe9-0d01dddbd58f@lexpan.law">
                  <meta http-equiv="Content-Type"
                    content="text/html; charset=UTF-8">
                  <p>I'm going repeat my comments on the MG-0 license
                    here since they are equally applicable to this
                    license (which appears to replicate the text of
                    MG-0, except for the addition of the conditions in
                    2.2):</p>
                  <p>1. The disclaimers are not made "conspicuous" as
                    that term is defined in UCC 2-316: <a
                      class="moz-txt-link-freetext"
                      href="https://www.law.cornell.edu/ucc/2/2-316"
                      moz-do-not-send="true">https://www.law.cornell.edu/ucc/2/2-316</a>
                    That has been interpreted as requiring something
                    like ALL CAPS or bold, or a different color, or a
                    box (although the criteria changed in 2022). This
                    isn't necessarily a flaw (whether UCC is relevant to
                    open source licenses is an interesting question) but
                    the practice seems to be that most newer open source
                    licenses try to adhere to this requirement (most by
                    using ALL CAPS since that tends to be the only way
                    to do this with .txt files or ASCII -- which
                    non-lawyers tend to dislike because they interpret
                    it as screaming without understanding why it's done
                    that way).<br>
                    <br>
                    2. I find the way the grants are structured
                    sub-optimal in the way that it handles the right of
                    performance under copyright law. Rather than being
                    in the grant, it is subsumed into the definition of
                    "Distribution/Distribute" and then grants a right to
                    Distribute. All rights are granted (which is good,
                    that way you don't have to rely on implied grants)
                    but you do need to dig into the definitions to get
                    there.</p>
                  <p>As to the Attribution version of the license, my
                    only comment is this license requires in Section
                    2.2(i) that a copy of the license be provided. This
                    is a fairly common provision of many so called
                    "permissive" or non-copyleft licenses although I've
                    always wondered what value this requirement
                    provides, given that this license is intended (I
                    believe) to be non-copyleft.</p>
                  <p>Otherwise, this license seems OK.</p>
                  <p>McCoy</p>
                  <p>[in my personal capacity and not as a member of the
                    board]</p>
                  <div class="moz-cite-prefix">On 6/18/2025 2:31 AM,
                    Moming Duan wrote:<br>
                  </div>
                  <blockquote type="cite"
cite="mid:883778D3-40CF-4DCF-B65F-AD07D9F427AD@gmail.com">
                    <meta http-equiv="content-type"
                      content="text/html; charset=UTF-8">
                    Dear OSI Community,
                    <div><br>
                    </div>
                    <div><br>
                    </div>
                    <div><span style="caret-color: rgb(0, 0, 0);">Following
                        our previous discussions in May, I have made
                        further revisions to the ModelGo </span>Attribution<span
                        style="caret-color: rgb(0, 0, 0);"> License
                        (MG-BY-2.0). I am submitting this updated
                        version for OSI review via this email. The
                        license text is attached.</span></div>
                    <div><br>
                    </div>
                    <div>
                      <div style="caret-color: rgb(0, 0, 0);"><font
                          color="#ff0000">\u2014\u2014\u2014\u2014\u2014\u2014 Major Updates to
                          Previous Submission</font></div>
                      <div style="caret-color: rgb(0, 0, 0);"><font
                          color="#ff0000"><br>
                        </font></div>
                      <div style="caret-color: rgb(0, 0, 0);">
                        <li><font color="#ff0000">Removes restrictions
                            on model output.</font></li>
                        <li><font color="#ff0000">Revises the
                            termination clause to provide for automatic
                            termination.</font></li>
                        <li><font color="#ff0000">Adds more explicit
                            granting of rights in Section 2.1. </font></li>
                        <li><font color="#ff0000">Narrows the definition
                            of \u201cDerivative Materials\u201d by including the
                            phrase: \u201cin order to replicate, approximate,
                            or otherwise achieve functional behavior
                            that is similar to the Model.\u201d </font></li>
                        <li><font color="#ff0000">Removes \u201cDerivative
                            Materials\u201d in Section 5: \u201cNothing in this
                            License permits You to modify this License
                            as applied to the Licensed Materials.\u201d </font></li>
                        <li><font color="#ff0000">Fixes typos and
                            formatting issues.</font></li>
                      </div>
                    </div>
                    <div><br>
                    </div>
                    <div><span style="caret-color: rgb(0, 0, 0);">\u2014\u2014\u2014\u2014\u2014\u2014 </span><span
                        style="caret-color: rgb(0, 0, 0);">License </span><span
                        style="caret-color: rgb(0, 0, 0);">Introduction</span></div>
                    <div><b><br>
                      </b></div>
                    <div><b>License Name</b>:<span
                      class="Apple-tab-span" style="white-space: pre;">          </span><span
                        style="caret-color: rgb(0, 0, 0);">ModelGo </span>Attribution<span
                        style="caret-color: rgb(0, 0, 0);"> License</span></div>
                    <div><span style="caret-color: rgb(0, 0, 0);"><b>Version</b>: <span
                        class="Apple-tab-span" style="white-space: pre;">                                </span>2.0</span></div>
                    <div><font><b>Short Identifier: <span
                          class="Apple-tab-span"
                          style="white-space: pre;">               </span></b>MG-BY-2.0</font></div>
                    <div><b style="caret-color: rgb(0, 0, 0);">Copyleft:</b><span
                      class="Apple-tab-span"
style="caret-color: rgb(0, 0, 0); font-weight: bold; white-space: pre;">                   </span><span
                        style="caret-color: rgb(0, 0, 0);">No</span></div>
                    <div><b>Legacy or New</b>: <span
                      class="Apple-tab-span" style="white-space: pre;">          </span>New
                      License</div>
                    <div><b>Drafted By Lawyer</b>: <span
                      class="Apple-tab-span" style="white-space: pre;">  </span>Yes, Rajah
                      &amp; Tann Singapore LLP</div>
                    <div><b>Approved or <span
                          style="caret-color: rgb(0, 0, 0);">Used</span> by
                        Projects</b>: <span class="Apple-tab-span"
                      style="white-space: pre;">   </span>No</div>
                    <div><br>
                    </div>
                    <div><b>License URL</b>:<span class="Apple-tab-span"
                      style="white-space: pre;">                           </span><a
                        href="https://ids.nus.edu.sg/modelgo-mg-by.html"
                        moz-do-not-send="true"
                        class="moz-txt-link-freetext">https://ids.nus.edu.sg/modelgo-mg-by.html</a></div>
                    <div><b>Introduction and Video</b>:<span
                      class="Apple-tab-span" style="white-space: pre;">  </span><a
                        href="https://www.modelgo.li/"
                        moz-do-not-send="true"
                        class="moz-txt-link-freetext">https://www.modelgo.li/</a></div>
                    <div><br>
                    </div>
                    <div><b>Overview</b>:</div>
                    <div><br>
                    </div>
                    <div>ModelGo Attribution License Version 2.0
                      (MG-BY-2.0) is a new license designed for
                      publishing models (typically neural networks like
                      Llama2, DeepSeek). It is one of the variants in
                      the ModelGo License family. MG-BY-2.0 is the a
                      permissive license in the ModelGo family,
                      requiring that the original license <font
                        color="#ff0000">and attribution</font> be
                      provided when distributing the original Licensed
                      Materials or Derivative Materials (<span
                        style="caret-color: rgb(0, 0, 0);">Licensed
                        Materials and </span><span
                        style="caret-color: rgb(0, 0, 0);">Derivative
                        Materials are</span><span
                        style="caret-color: rgb(0, 0, 0);"> </span>defined
                      in Clause 1). <font color="#ff0000">A statement of
                        modification is required, if applicable.</font></div>
                    <div><font color="#ff0000">(Red content represents
                        the differences from MG0-2.0 license)</font></div>
                    <div><br>
                    </div>
                    <div><b>Complies with OSD:</b></div>
                    <div><b><br>
                      </b></div>
                    <div>OSD 3 Derived Works \u2014 MG-BY-2.0 <span
                        style="caret-color: rgb(0, 0, 0);"> </span><span
                        style="caret-color: rgb(0, 0, 0);">Clause 2.1
                        (a) grants copyright and patent rights to create
                        derivatives.</span></div>
                    <div>OSD 5 and OSD 6 \u2014 No discrimination clause is
                      included in MG-BY-2.0.</div>
                    <div>OSD 9 License Must Not Restrict Other Software
                      \u2014 No such restriction is included in MG-BY-2.0.</div>
                    <div><br>
                    </div>
                    <div><b>The Gap to Fill:</b></div>
                    <div>Model sharing is very common on the web, with
                      over 1.4 million models currently listed on
                      Hugging Face (<a
                        href="https://huggingface.co/models"
                        moz-do-not-send="true"
                        class="moz-txt-link-freetext">https://huggingface.co/models</a>).
                      However, most of these models are not properly
                      licensed. When publishing their models, developers
                      typically choose from three main options (as seen
                      in the model license tags on the Hugging Face
                      website):</div>
                    <div><br>
                    </div>
                    <div>
                      <ul class="MailOutline">
                        <li>OSS licenses, e.g., Apache-2.0, MIT</li>
                        <li>Open responsible AI licenses (OpenRAILs),
                          e.g., CreativeML-OpenRAIL-M, OpenRAIL++</li>
                        <li>Proprietary Licenses, e.g., Llama2, Llama3</li>
                      </ul>
                    </div>
                    <div><br>
                    </div>
                    <div>However, not all licenses are well-suited for
                      model publishing.</div>
                    <div><br>
                    </div>
                    <div><b>Why not use OSS licenses? </b></div>
                    <div>Traditional OSS licenses lack clear definitions
                      regarding machine learning concepts, such as
                      Models, Output, and Derivatives created through
                      knowledge transfer. This ambiguity can result in
                      certain ML activities (e.g., Distillation,
                      Mix-of-Expert) being beyond the control of the
                      model owner.</div>
                    <div><br>
                    </div>
                    <div><b>Why not use OpenRAILs? </b></div>
                    <div>Recently, Responsible AI Licenses (<a
                        href="https://www.licenses.ai/"
                        moz-do-not-send="true"
                        class="moz-txt-link-freetext">https://www.licenses.ai/</a>)
                      have been widely advocated to govern AI
                      technologies, aiming to restrict unlawful and
                      unethical uses of models. While I acknowledge the
                      growing need for such governance, these
                      copyleft-style restrictions do not comply with the
                      OSD and may cause incompatibility with licenses
                      like GPL-3.0. Another concern is that these
                      behavioral restrictions may proliferate within the
                      AI model ecosystem, increasing the risk of license
                      breaches.</div>
                    <div><br>
                    </div>
                    <div><b style="caret-color: rgb(0, 0, 0);">Why not
                        use Llama2 or Llama3 Licenses?</b></div>
                    <div><font>These licenses are proprietary licenses
                        that are not reusable. </font>Furthermore, they
                      include exclusive terms such as "You will not use
                      the Llama Materials or any output or results of
                      the Llama Materials to improve any other large
                      language model" and copyleft-style behavioral
                      restrictions.</div>
                    <div><br>
                    </div>
                    <div>In fact, the dilemma in current model
                      publishing is the lack of a general-purpose
                      license for model developers. Additionally, since
                      no single license meets diverse model publishing
                      needs, some developers resort to using CC licenses
                      with different elements. However, CC licenses are
                      ill-suited for this purpose as they do not grant
                      patent rights. This motivated the drafting of
                      ModelGo License family, which provides different
                      licensing elements similar to CC but specifically
                      designed for model publishing.</div>
                    <div><br>
                    </div>
                    <div><b>Comparison with Existing OSI-Approved
                        Licenses:</b></div>
                    <div>Since I could not find an OSI-approved model
                      license, I can only compare MG-BY-2.0 with one
                      similar OSS license \u2014 Apache-2.0</div>
                    <div>
                      <div><br>
                      </div>
                      <div>
                        <li style="caret-color: rgb(0, 0, 0);">MG-BY-2.0
                          defines licensed materials and derivative
                          works differently from Apache-2.0, tailoring
                          them to models.</li>
                        <li style="caret-color: rgb(0, 0, 0);">MG-BY-2.0
                          can govern the remote access (e.g., chatbot)
                          scenario.</li>
                      </div>
                    </div>
                    <div><br>
                    </div>
                    <div>If further comparisons or supporting evidence
                      are needed to strengthen my claims, please let me
                      know. I am more than willing to engage in further
                      discussions with the OSI community about this
                      license and contribute to promoting standardized
                      model publishing. <span
                        style="caret-color: rgb(0, 0, 0);">\U0001f917</span></div>
                    <div><br>
                    </div>
                    <div><br>
                    </div>
                    <div>Best,</div>
                    <div>Moming</div>
                    <div><br>
                    </div>
                    <br>
                    <fieldset class="moz-mime-attachment-header"></fieldset>
                    <meta http-equiv="content-type"
                      content="text/html; charset=UTF-8">
                    <br>
                    <fieldset class="moz-mime-attachment-header"></fieldset>
                    <pre class="moz-quote-pre" wrap="">_______________________________________________
The opinions expressed in this email are those of the sender and not necessarily those of the Open Source Initiative. Communication from the Open Source Initiative will be sent from an opensource.org email address.

License-review mailing list
<a class="moz-txt-link-abbreviated moz-txt-link-freetext"
                    href="mailto:License-review@lists.opensource.org"
                    moz-do-not-send="true">License-review@lists.opensource.org</a>
<a class="moz-txt-link-freetext"
href="http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org"
                    moz-do-not-send="true">http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org</a>
</pre>
                  </blockquote>
                  <br>
                  <fieldset class="moz-mime-attachment-header"></fieldset>
                  <pre class="moz-quote-pre" wrap="">_______________________________________________
The opinions expressed in this email are those of the sender and not necessarily those of the Open Source Initiative. Communication from the Open Source Initiative will be sent from an opensource.org email address.

License-review mailing list
<a class="moz-txt-link-abbreviated moz-txt-link-freetext"
                  href="mailto:License-review@lists.opensource.org"
                  moz-do-not-send="true">License-review@lists.opensource.org</a>
<a class="moz-txt-link-freetext"
href="http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org"
                  moz-do-not-send="true">http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org</a>
</pre>
                </blockquote>
                <br>
              </div>
              _______________________________________________<br>
              The opinions expressed in this email are those of the
              sender and not necessarily those of the Open Source
              Initiative. Communication from the Open Source Initiative
              will be sent from an opensource.org email address.<br>
              <br>
              License-review mailing list<br>
              <a class="moz-txt-link-abbreviated" href="mailto:License-review@lists.opensource.org">License-review@lists.opensource.org</a><br>
<a class="moz-txt-link-freetext" href="http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org">http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org</a><br>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
  </body>
</html>