Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

The elephant in the room there is that if you allow AI contributions you immediately have a licensing issue: AI content can not be copyrighted and so the rights can not be transferred to the project. At any point in the future someone could sue your project because it turned out the AI had access to code that was copyrighted and you are now on the hook for the damages.

Open source projects should not accept AI contributions without guidance from some copyright legal eagle to make sure they don't accidentally exposed themselves to risk.



Well, after today's incidents I decided that none of my personal output will be public. I'll still license them appropriately, but I'll not even announce their existence anymore.

I was doing this for fun, and sharing with the hopes that someone would find them useful, but sorry. The well is poisoned now, and I don't my outputs to be part of that well, because anything put out with well intentions is turned into more poison for future generations.

I'm tearing the banners down, closing the doors off. Mine is a private workshop from now on. Maybe people will get some binaries, in the future, but no sauce for anyone, anymore.


Yeah I’d started doing this already. Put up my own Gitea on my own private network, remote backups setup. Right now everything stays in my Forge, eventually I may mirror it elsewhere but I’m not sure.


this is exactly what I've been doing for the past 3 years

and my internet comments are now ... curated in such a way that I wouldn't mind them training on them


Well, well, well, seems you're onto something here.


You and many more like you.


Damn, the Dark Forest is already coming for open source

https://maggieappleton.com/ai-dark-forest

tl;dr: If anything that lives in the open gets attacked, communities go private.


> AI content can not be copyrighted and so the rights can not be transferred to the project. At any point in the future someone could sue your project because it turned out the AI had access to code that was copyrighted and you are now on the hook for the damages.

Not quite. Since it has copyright being machine created, there are no rights to transfer, anyone can use it, it's public domain.

However, since it was an LLM, yes, there's a decent chance it might be plagiarized and you could be sued for that.

The problem isn't that it can't transfer rights, it's that it can't offer any legal protection.


So far, in the US, LLM output is not copyrightable:

https://www.congress.gov/crs-product/LSB10922


Yes, I said that. That doesn't mean that the output might not be plagiarized. I was correcting that the problem wasn't about rights assignment because there are no rights to assign. Specifically, no copyrights.


> Since it has copyright being machine created, there are no rights to transfer, anyone can use it, it's public domain.

Maybe you meant to include a "doesn't" in that case?


should have said "Since it has no copyright"


Sorry, this doesn't make sense to me.

Any human contributor can also plagiarize closed source code they have access to. And they cannot "transfer" said code to an open source project as they do not own it. So it's not clear what "elephant in the room" you are highlighting that is unique to A.I. The copyrightability isn't the issue as an open source project can never obtain copyright of plagiarized code regardless of whether the person who contributed it is human or an A.I.


a human can still be held accountable though, github copilot running amock less so


If you pay for Copilot Business/Enterprise, they actually offer IP indemnification and support in court, if needed, which is more accountability than you would get from human contributors.

https://resources.github.com/learn/pathways/copilot/essentia...


I think that they felt the need to offer such a service says everything, basically admitting that LLMs just plagiarize and violate licenses.


9 lines of code came close to costing Google $8.8 billion

how much use do you think these indemnification clauses will be if training ends up being ruled as not fair-use?


Are you concerned that this will bankrupt Microsoft?


I think they're afraid they will have to sue Microsoft to get them to abide by the promise to come to their defense in another suit.


be nice, wouldn't it?

poetic justice for a company founded on the idea of not stealing software


That covers any random contribution claiming to be AI?


Their docs say:

> If any suggestion made by GitHub Copilot is challenged as infringing on third-party intellectual property (IP) rights, our contractual terms are designed to shield you.

I'm not actually aware of a situation where this was needed, but I assume that MS might have some tools to check whether a given suggestion was, or is likely to have been, generated by Copilot, rather than some other AI.


Human beings can create copyrightable code.

As per the US Copyright Office, LLMs can never create copyrightable code.

Humans can create copyrightable code from LLM output if they use their human creativity to significantly modify the output.


AI code by itself cannot be protected. However the stitching together of AI output and curation of outputs creates a copyright claim.


You may indeed have a licensing issue... but how is that going to be enforced? Given the shear amount of AI generated code coming down the pipes, how?


I doubt it will be enforced at scale. But, if someone with power has a beef with you, it can use an agent to search dirt about you and after sue you for whatever reason like copyright violation.


It will be enforced by $BIGCORP suing $OPEN_SOURCE_MAINTAINER for more money than he's got, if the intent is to stop use of the code. Or by $BIGCORP suing users of the open source project, if the goal is to either make money or to stop the use of the project.

Those who lived through the SCO saga should be able to visualize how this could go.


If you were foolish enough to send your code to someone else's LLM service, they know exactly where you used their output.

If they wanted to, they could take that output and put you out of business because the output is not your IP, it can be used by anybody.


It will be enforced capriciously by people with more money than you and a court system that already prefers those with access and wealth.


> At any point in the future someone could sue your project because it turned out the AI had access to code that was copyrighted and you are now on the hook for the damages.

So it is said, but that'd be obvious legal insanity (i.e. hitting accept on a random PR making you legally liable for damages). I'm not a lawyer, but short of a criminal conspiracy to exfiltrate private code under the cover of the LLM, it seems obvious to me that the only person liable in a situation like that is the person responsible for publishing the AI PR. The "agent" isn't a thing, it's just someone's code.


That's why all large-scale projects have Contributor License Agreements. Hobby/small projects aren't an attractive legal target--suing Bob Smith isn't lucrative; suing Google is.


You might find that the AI accepts that as a valid reason for rejecting the PR.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: