| 1. _Material_ licensing fees for use of protected content. NYT will have grounds for punitive damages (over 100K per work per infringing use), enough money to make OpenAI and even Microsoft come to the negotiating table. 2. Open door for everyone else whose content has been copied to demand payment. 3. Japan aside*, I'm not sure how a decisive victory for NYT fails to affect companies in other countries. Other countries will certainly take a US opinion into account. Furthermore, nothing whatsoever would prevent NYT from suing for infringement in France, Germany, etc. A number of European countries are _much_ more comfortable with compulsory licensing than the US is.** I will not be surprised at all if Germany, Denmark, Sweden etc impose an AI-training license that companies have to pay to train or even use AI models. Could be based on employees, users, revenue, etc. They've been doing this for decades for other forms of mechanical reproduction, so it seems a reasonable path forward. * I've seen reports that Japan has declared that it will not enforce copyright against organizations for training models on copyright-protected works. I don't know whether this means that _everything_ is fair game in Japan-- it could mean that training won't be enforced, but commercial use of models trained on protected works could be. I think it's somewhat harder to argue against training a model than it is against leveraging a trained model for commercial purposes. ** AFAIK, the only use for compulsory licensing in the US is musical performance. 4. "Surely one can't get licenses"... collective licensing would be a path forward for which there's a little more history in the US than there is for compulsory licensing. In a collective licensing model, OpenAI/Microsoft/Company X would be able to purchase a license priced by revenue/employees/users etc, and funds would be distributed to participating copyright holders. For that to work, you need some kind of agency that has a ton of significant copyright holders. If some agency were able to get to a critical mass , that could happen without Congressional action. Congress could also step into the problem in a couple of ways. They could wave a wand and make the problem go away for OpenAI/Microsoft/everyone else, or they could explicitly require some kind of license. I'd be shocked if Congress handled this productively _at all_; these issues are incredibly fraught in the best of times, Congress has been terrible for years at any kind of precise technical work, and the current Congress is uniquely incompetent. I expect that if Congress accomplishes anything on this topic in the next couple of years, it will turn out to be awful for everyone. |