AI Companies and Creators Clash Over Copyright Laws for Text Generation
-
AI companies like OpenAI and Google argue training models on copyrighted content is like a person learning from reading books. But creators counter AI directly copies their work.
-
Determining copyright infringement will likely be decided case-by-case. Challenges include proving copying happened and that outputs are similar enough.
-
EU copyright law may favor rightsholders more than US "fair use", with a directive allowing creators to opt-out of text and data mining.
-
Transparency rules in the EU's new AI Act could help rightsholders, but may not change things fundamentally.
-
Some experts believe AI companies have "already won" by starting model training before comprehensive regulations were in place.