There is simply no way that is true. First, the legal arguments are dodgy:

1. There is a good chance that LLMs are sufficiently transformative that courts will decide they don’t infringe the copyright of their sources.
2. Even if not, hosts will get DMCA-style safe harbour protection and the most they’ll be liable for is takedown requests, which is a major thing for any large host.

Second, there is no fucking way a western court is going to tell every big tech company they have to delete 99% of the code that was written since the start of the year, even if the law as written literally said verbatim, “use of any LLM output for any purpose is breach of copyright” because it doesn’t take a conspiracy theorist to realise that it’s politically impossible.

You may not like that, but it means that, again, Codeberg is doing this based on feels.

Not true, this does not occur frequently. This study and software LiCoEval from 2024 found 0.88% to 2.01% of code “strikingly similar to existing open-source implementations”. Afaik this is mostly textbook examples, snippets from stack overflow snippets or common github repositories, often replicated api examples and language boilerplate. How you prompt and refine also matters, and for generating novel code or business logic the LLM simply cannot use memorized snippets.

Presumably since then LLMs have worked to reduce that number of memorized code. Since LLMs cannot memorize all their training data, that number is limited. LiCoEval can find the often memorized examples and train to remove them, or suppress them, or they find other ways to reduce direct reproduction from memorization. For example it would be possible to do what malus.sh does with all the training data. Then it cannot memorize copyrighted code.

So for a model that came out 2026 this already small number might not be that relevant anymore.