Fair use benefits everyone
Various organizations, creators, and experts have voiced support for fair use as it applies to generative AI training, arguing that it is transformative, promotes lawful innovation, and does not cause cognizable market harm. Amicus briefs and statements emphasize that restricting fair use would hinder scientific progress, restrict accessibility tools, and negatively impact independent creators and researchers who rely on AI systems. Open article →
Notable comments
The discussion broadly supports fair use for AI training on transformative/beneficial grounds, while several commenters sharply distinguish it from wholesale memorization or monetizing “automation of fair use,” and dispute claims by pointing to concrete litigation and downstream distillation concerns.
HN · simonw Argues the NYT suit is not just about training-in-general but alleges memorization, citing examples where OpenAI models can reproduce NYT articles from weights.
HN · erikschoster Raises a boundary case: while excerpting a comic could be fair use, building a business that automates fair use is claimed to be qualitatively different (likely not fair use).
HN · simonw Pushes back on a prior framing by noting Bartz v. Anthropic was not “about fair use” per se, clarifying that the judge found training on scanned, purchased books fair use and giving context via a linked analysis.