LLMs and the Law: training doesn't make a model a derivative work, FAQ explains
esrtweet · x · 2026-10-26
A detailed FAQ (assisted by ChatGPT 6.0, current as of 2026-10-25) addresses common claims about LLMs and US copyright law.
- An LLM is not a derivative work of its training set merely because it was trained on it; derivative works must "recast, transform, or adapt" preexisting works and courts generally require incorporation of protected expression, not just learning facts, ideas, statistical relationships, vocabulary, or style (17 U.S.C. §101, §102(b), §106(2)).
- A model could theoretically infringe if its weights encode recoverable protected expression; as of October 2026, no such case has been reported.
More from Safety
- Several well-known SaaS hacked last week, AI API keys leaked en masse and abused — draginol · 2026-10-26
- ChatGPT adds PIN-locked private chats — but forgetting the PIN erases them for good — ThereWas · 2026-10-26
- Claude's newer models refuse reverse engineering tasks where older ones complied — Toiling-Donkey · 2026-10-26
- Georgia officials: Anthropic-linked bot scraped public voter data, didn't touch registration system — asusarla · 2026-10-26
- PHP WASM servers show no sign of PHP2SHELL impact so far, patch in progress — MickeySteamboat · 2026-10-26
- Apple's week-long curl/git CA bundle lag sparks developer security debate — kingslyj · 2026-10-26