US government sides with OpenAI on issue of training LLMs on copyrighted material
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
The US government filed a 20-page brief arguing that unlicensed use of copyrighted material for LLM training qualifies as fair use. This removes near-term legal risk for production deployments and lowers compliance costs, but it also locks in the current data-scraping model—if courts later reverse, retroactive licensing or model retraining could become mandatory.
The federal government has now filed a brief backing the fair-use defense for training LLMs on unlicensed copyrighted material, adding executive-branch weight to a legal trend (Alsup's Anthropic ruling) that treats training itself as transformative and legal—while penalizing only pirated acquisition. Practical upshot: the training-legality risk to your model pipeline keeps shrinking, but data provenance is the exposure that matters—source your training corpora through legitimately acquired channels, because how you obtained the data, not that you trained on it, is what draws liability.