Been doing a bunch of thinking about LLM code, the (known) uncopyrightability thereof, and the impact that may have on licences and rights holders downstream.

If your open-source project contains LLM code, how does that impact your chosen licence? If you can’t hold copyright on it, can you assert your other rights under the licence model?

What about in a dependency?

I feel like the definitive answer is beyond me right now.

I clearly remember the panic about GPL code at Microsoft in the ‘00s - just *seeing* GPL code as a techie set off all manner of alarm bells in management, lest the GPL “pollute” MS’s closed code. Very Serious Memos were circulated.

The LLM situation seems to have parallels with that.

We’ll probably work it out. MS did. But it warrants attention.

@drunkenmadman except it's worse woth #AIslop / #SlopCode because that shit is unmaintainable
@drunkenmadman short answer: you can’t copyright AI output. However it won’t invalidate a previously copyrighted work. So we can have software that is partly copyright and partly not. Wait for Microsoft to try arguing that in court
@drunkenmadman Seems to me that the fact that most if not all LLM-generated code contains code stolen from other sources would be a bigger concern...?