News for hackers. Stuff that matters.
Re:GitHub Copilot is generally available by dahart on Tuesday June 21, 2022 @10:33PM (#31831446)
> Someone notices this and sues the commercial project for breaking GPL v3.

I know what you mean, but silly nit pick since you mentioned “commercial” twice - GPL v3 does not prevent commercial use, it only requires copies to be open source. For someone to notice the project has copied code and not be inside the company, the code would (probably) have to be open source. So, this hypothetical is less likely to happen than your comment makes it sound.

A little further off topic, but amusing to me, is that the US government defines “commercial” software to be any software that has a license other than public domain. Free and open source software, such as GPL v3, is still “commercial” because it is licensed to the public https://www.acquisition.gov/far/part-2#FAR_2_101

More on-topic now, a small single function accidentally copied from an open source project by automated software might be considered fair use by US copyright law. https://www.copyright.gov/fair-use/more-info.html

(Edit) Oh yeah, and I just remembered that GitHub’s Terms already carve a necessary exception to whatever license you use in your project, to allow Github to host & display your code. I assume those terms already include some CoPilot coverage…? If not, and if they aren’t legally covered already (which I bet they are), then they could change the terms to stipulate that hosting code on GitHub bars people from suing over incidental amounts of automated copying. Main point here being that the GPLv3 license on your project is neither the only nor the primary license governing GitHub’s relationship with your code.

Re:GitHub Copilot is generally available by account42 on Wednesday June 22, 2022 @04:00AM (#31833206)
> I know what you mean, but silly nit pick since you mentioned “commercial” twice - GPL v3 does not prevent commercial use, it only requires copies to be open source. For someone to notice the project has copied code and not be inside the company, the code would (probably) have to be open source. So, this hypothetical is less likely to happen than your comment makes it sound.

The are plenty of source available or open core projects where use of GPL-ed code would both be visible and incompatible with the licensing.

> Oh yeah, and I just remembered that GitHub’s Terms already carve a necessary exception to whatever license you use in your project, to allow Github to host & display your code. I assume those terms already include some CoPilot coverage…?

Letting github host and display the code is compatible with open source licenses but is very much different from letting third parties incorporate that code into non-open codebases.

> If not, and if they aren’t legally covered already (which I bet they are), then they could change the terms to stipulate that hosting code on GitHub bars people from suing over incidental amounts of automated copying. Main point here being that the GPLv3 license on your project is neither the only nor the primary license governing GitHub’s relationship with your code.

GitHub TOS can only possibly give them authorization from those directly uploading code to GitHub. They don't give github any additional license for code that was uploaded by a someone else (for example a mirror bot) becose that someone does not have the rights to give out that license.

And even if they write in their TOS that they can do whatever they want it does not mean that they can actually legally do whatever they want - even moreso when retroactively changing those terms.

Re:GitHub Copilot is generally available by dahart on Wednesday June 22, 2022 @10:34AM (#31836524)
> Letting github host and display the code is compatible with open source licenses but is very much different from letting third parties incorporate that code into non-open codebases.

Agree completely. It’s still a fact that GitHub’s Terms provide a separate license for Github. See section D.4 https://docs.github.com/en/site-policy/github-terms/github-t...

> GitHub TOS can only possibly give them authorization from those directly uploading code to GitHub. They don’t give github any additional license for code that was uploaded by a someone else (for example a mirror bot) becose that someone does not have the rights to give out that license.

GitHub’s terms already require that uploaders have copyright authorization, or that machine uploaders are doing automated tasks exclusively. Letting a machine upload someone else’s new and copyright code that the account owner doesn’t have copyrights to appears to violate GitHub’s terms. “the owner of the Account is ultimately responsible for the machine's actions” https://docs.github.com/en/site-policy/github-terms/github-t...

> even if they write in the TOS that they can do whatever they want it does not mean that they can actually legally do whatever they want

Yes, correct, I completely agree. Asking for people to agree to ‘indemnify and hold harmless’ over the site’s features is standard language and not really a stretch, it doesn’t amount to Github doing whatever they want. That language is already in the terms.

What is the summary of your comment, what are you trying to say at a high level? We can debate the fine points, but if you are trying to say that my speculative suggestion was crappy and Github has other ways to make copilot legal, then I agree. If you’re trying to say that Github has no legal way to make copilot available, then I disagree.

There is already language in the terms that might cover copilot. Section D.4 I linked to above includes this:

“This license [that you grant us] includes the right to do things like copy it [your content] to our database and make backups; show it to you and other users; parse it into a search index or otherwise analyze it on our servers; share it with other users; and perform it, in case Your Content is something like music or video.“

The Copilot FAQ also mentions they have IP filters and actively prevent reciting large portions of anyone’s code, it explicitly mentions a threshold of 150 characters.