> I want to write less, just knowing that LLM models are going to be trained on my code is making me feel more strongly than ever that my open source contributions will simply be stolen. Am I wrong to feel this? Is anyone else concerned about this?
I don't think it's wrong, but misdirected maybe. What do you that someone can "steal" your open source contributions? I've always released most of my code as "open source", and not once has someone "stolen" it, it still sits on the same webpage where I initially published it, decades ago. Sure, it's guaranteed ingested into LLMs since long time ago, but that's hardly "stealing" when the thing is still there + given away for free.
I'm not sure how anyone can feel like their open source code was "stolen", wasn't the intention in the first place that anyone can use it for any purpose? That's at least why I release code as open source.
"Open Source" does not equal "No terms on how to share and use the code". Granted, there are such licenses but afaik the majority requires attribution at the minimum.
Then I'd say they're "breaking the license", not "stolen your project", but maybe I'm too anal about the meaning of words.
Yeah, fair, I could have been clearer. But yes, that is what I meant: breaking the license.
I’m unaware of any mainstream Open Source licenses that forbid training an AI model on the work. Are you using one?
[A]GPL is viral, so the derived code must use the same license. People that like that license care a lot about that.
On the other side BSD0 is just a polite version of WTFPL, and people that like it doesn't care about what you do with the code.
And I mostly use MIT, which requires attribution. Does that mean when people use my code, without attribution me, that they're "stealing my code"? I would never call it that, I'd say they're "breaking the license", or similar.
The MIT license doesn’t require attribution for “using...code.” It reads as follows:
> 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, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
> The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
> THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
The operative language here is “all copies or substantial portions of the Software.” LLMs, with rare exceptions, don’t retain copies or substantial portions of the software it was trained on. They’re not libraries or archives. So it’s unclear to me how training an AI model with an MIT-licensed project could violate the license.
(IAAL and this is my personal analysis, not legal advice.)
I think the GP said "use" in the programer sense, i.e. ctr-C&ctr-V into your program. Not in the normal sense, i.e. double click on the icon. So I guess we all agree.