It’s interesting to listen to people’s reactions to Judge Bibas’s stunning reversal in ruling that ROSS Intelligence’s training of its AI model on copyrighted Westlaw headnotes for uncopyrightable judicial decisions was not transformative or different in purpose, and, ultimately, not a fair use.
Jason Calacanis of the All In podcast interpreted the decision as suggesting that OpenAI might lose its fair use defense against the New York Times. But, inexplicably, Calacanis puts the odds of OpenAI losing its fair use defense at only 5 to 10%. (Also interesting that Calacanis only mentioned OpenAI, when Meta, Google, Microsoft, Anthropic, Perplexity, Cohere, Midjourney, and other AI startups also face copyright lawsuits raising the same issue.)
Given that the legal issue is novel, with one district court ruling against fair use in AI training, the odds against fair use seem like they should be higher--at least 50%, or no better than a toss-up. Underestimating the odds of copyright liability in Silicon Valley might be a problem itself.
We asked ChatGPT and its answer was closer to ours: 60% odds the Supreme Court finds fair use.
While it’s impossible to predict the future with certainty—and I’m not offering legal advice—if I had to assign odds based on existing precedent and trends in the Court’s reasoning, I might put it at roughly a 60% chance that the Supreme Court would rule a company’s use of copyrighted works to train AI models qualifies as fair use.
This estimation hinges on several factors:
- Transformative Use: Past decisions suggest that if the new use is sufficiently transformative—adding new expression, meaning, or value—it may lean toward fair use.
- Precedents: Cases like Google Books and other related copyright disputes have shown some judicial leniency when it comes to digital transformation.
- The Evolving Tech Landscape: The Court is increasingly aware of the impact of digital innovation, though the exact legal framework is still developing.
Remember, this is a speculative odds estimate rather than a legal forecast.
No doubt many people (especially ones who dislike all things AI) put the odds much worse for fair use–some perhaps even suggesting that the odds are 100% that fair use is rejected. Or no chance.
That seems too high in the opposite direction, given the novel legal issue raised and the existence of two technological fair use decisions by the Supreme Court (Sony and Google), even excluding the circuit court decisions recognizing technological fair uses in other contexts. Indeed, if we gave odds on the fair use defense before either Sony or Google was decided, the odds against fair use probably would have been quite high, more than 50%. What the Supreme Court will say, ultimately, is too hard to predict.
And, yes, we should fully expect this novel legal issue of fair use, from one or more of the AI copyright lawsuits, will one day reach the Supreme Court. The issue is too important to United States for the Supreme Court to ignore it.
Related stories