[License-discuss] Curious about derived works and AI...

Bruce Perens bruce at perens.com
Thu May 30 20:40:43 UTC 2024


The path, as Shuji-san implied, is by using contract terms, and it's going
to be a little different from Open Source. I don't know if anyone has a
lawyer-approved license ready for you. You can look at what I'm working
toward at postopen.org . The license there has contract terms that I hope
will work, but are not through legal review yet.

On Thu, May 30, 2024, 12:46 Miles Georgi <azimux at gmail.com> wrote:

> Thanks so much for the insight, Bruce, Shuji, and Josh.
>
> Sounds like I find myself at a bit of a murky point in history for open
> source project licensing/copyright. I was wondering if maybe weak copyleft
> licenses might help over permissive licenses but honestly seems like that's
> irrelevant. I'd be eager to take a small hit in adoption if I felt like I
> could protect contributors from such licensing/copyright circumvention. But
> it all sounds so uncharted that I don't see an obvious path.
>
> One thing I am considering is starting to use SPDX headers in source-code
> files so at least AI systems are aware that the input is licensed and
> copyrighted, though doesn't seem likely to have any practical impact. But I
> suppose easy enough to automate and probably doesn't hurt much.
>
> But yeah, sounds like, for now, I just have to make a choice ignoring this
> concern.
>
> Thanks again!
>
> Cheers
>
> Miles
>
>
> On Tue, May 28, 2024 at 1:14 AM Bruce Perens via License-discuss <
> license-discuss at lists.opensource.org> wrote:
>
>> Hi Sado-san,
>>
>> It is unfortunate that Japan made the choice that it did, because it
>> makes Open Source software fair game for those who would profit from the
>> work of our developers without attribution or remuneration and entirely
>> outside of their license terms. I came to the conclusion that this could
>> not be fixed under my own rules in the OSD. This is one of the reasons I've
>> been working on Post-Open. At this time the Post-Open license requires
>> explicit consent for use in Japan (rather than tear-open or implicit
>> consent) because I felt that was necessary to work reliably within their
>> law.
>>
>>     Thanks
>>
>>     Bruce
>>
>> On Sat, May 25, 2024 at 5:31 PM Shuji Sado <shujisado at gmail.com> wrote:
>>
>>> In general, Japanese copyright law is considered to be one of the most
>>> compatible with machine learning in the world.
>>> Training AI with open source code is basically considered legal, and
>>> even for commercial software code, AI training is legal as long as it
>>> is not explicitly prohibited by contract.
>>>
>>> However, if the output generated by the AI is substantially identical
>>> to the original code, it will be considered copyright infringement of
>>> the original code.
>>> License washing is not permitted.
>>>
>>> The Agency for Cultural Affairs in Japan recently published an English
>>> PDF titled "General Understanding on AI and Copyright in Japan":
>>> https://www.bunka.go.jp/english/policy/copyright/pdf/94055801_01.pdf .
>>> Various industries in Japan discussed and submitted tens of thousands
>>> of public comments for the creation of this document.
>>> Although it is not a complete document, it provides an understanding
>>> of how Article 30-4 of the Copyright Act, which is seen as fully
>>> endorsing machine learning, actually functions.
>>>
>>>
>>> 2024/05/26 2:10 Bruce Perens via License-discuss
>>> <license-discuss at lists.opensource.org>:
>>> >
>>> > 1. Not in Japan, because they've decided to make their law that way.
>>> 2. It should be the case in most countries, but it is not so far because
>>> it's not literal copying and cases which are attempting to make the point
>>> that it is copying are still in litigation.
>>> >
>>> > On Sat, May 25, 2024, 06:55 Miles Georgi <azimux at gmail.com> wrote:
>>> >>
>>> >> Hi! If I open-source a project with a free license, if code from that
>>> project winds up being used as training data or prompt data to a
>>> code-generating AI to generate similar code, would that generated code be
>>> considered a derived work under any circumstances? And does that
>>> potentially depend on what license is chosen?
>>> >>
>>> >> I'm trying to choose a license for a project I wish to release and
>>> I'm assuming there's no way to protect against something like that but
>>> figured I'd ask.
>>> >>
>>> >> Cheers!
>>> >>
>>> >> Miles
>>> >> _______________________________________________
>>> >> The opinions expressed in this email are those of the sender and not
>>> necessarily those of the Open Source Initiative. Official statements by the
>>> Open Source Initiative will be sent from an opensource.org email
>>> address.
>>> >>
>>> >> License-discuss mailing list
>>> >> License-discuss at lists.opensource.org
>>> >>
>>> http://lists.opensource.org/mailman/listinfo/license-discuss_lists.opensource.org
>>> >
>>> > _______________________________________________
>>> > The opinions expressed in this email are those of the sender and not
>>> necessarily those of the Open Source Initiative. Official statements by the
>>> Open Source Initiative will be sent from an opensource.org email
>>> address.
>>> >
>>> > License-discuss mailing list
>>> > License-discuss at lists.opensource.org
>>> >
>>> http://lists.opensource.org/mailman/listinfo/license-discuss_lists.opensource.org
>>>
>>>
>>>
>>> --
>>> Shuji Sado
>>> Chairman, Open Source Group Japan
>>> https://opensource.jp/
>>> https://shujisado.com/
>>>
>>> _______________________________________________
>>> The opinions expressed in this email are those of the sender and not
>>> necessarily those of the Open Source Initiative. Official statements by the
>>> Open Source Initiative will be sent from an opensource.org email
>>> address.
>>>
>>> License-discuss mailing list
>>> License-discuss at lists.opensource.org
>>>
>>> http://lists.opensource.org/mailman/listinfo/license-discuss_lists.opensource.org
>>>
>>
>>
>> --
>> Bruce Perens K6BP
>> _______________________________________________
>> The opinions expressed in this email are those of the sender and not
>> necessarily those of the Open Source Initiative. Official statements by the
>> Open Source Initiative will be sent from an opensource.org email address.
>>
>> License-discuss mailing list
>> License-discuss at lists.opensource.org
>>
>> http://lists.opensource.org/mailman/listinfo/license-discuss_lists.opensource.org
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensource.org/pipermail/license-discuss_lists.opensource.org/attachments/20240530/bc5a5f99/attachment-0001.html>


More information about the License-discuss mailing list