John HardyCloudflare by vibe-cloning Next.js are laundering an open source piece of software and converting it into something with a different licence.
That is NOT in the spirit or intention of open source!
Leszek@jhlagado Just a reminder that the US Supreme Court decision only means there's no new copyrightable work created by a non-human program in the US legal system. It doesn't mean the original authors of the source work (whether it's training data or "inspiration" for a solution referred in the prompt) lose their copyright.
Sure. Are you meaning Vercel the original authors? Yes, the status of their work, of Next.js, is unchanged.
That goes for Anthropic, the authors of Claude code, too.
I'm only referring to the copyright status of the new software written by CloudFlare. The vibe coded copy of Next.js has no copyright protection or open source status.
@jhlagado I mean the output of LLMs might still have copyright (depending on interpretation in currently pending lawsuits), just not held by the person who ran the LLM/provided the prompt. The SC interpretation pivots on creativity requiring a human author, it says nothing about training data copyright being washed away.
I'm watching NYT vs. OpenAI closely since it's really close to the direct question if training an LLM on something launders away any copyright on the input data.
@makdaam The US Copyright Office has said that the existing laws are adequate and new laws are not needed. AI assisted software may be copyrightable but vibe coded software is not likely to be copyrightable.
https://www.congress.gov/crs-product/LSB10922
The question of training is less positive for property owners. There have been court cases where the verdict is that training in LLM on copyright material is considered "fair use" as long as a copy of the material used in the training was purchased by the LLM owner.
