December 27, 2023. The New York Times. filed a lawsuit against OpenAI, alleging that the company committed willful copyright infringement through its ChatGPT generative artificial intelligence tool. The Times claimed each that ChatGPT was illegally trained on vast amounts of text from its articles and that ChatGPT’s output contained language taken directly from its articles.
To treatment this, the Times demanded greater than just money: it asked a federal court to order the “destruction” of ChatGPT.
If granted, the request would force OpenAI to delete trained models of large languages equivalent to GPT-4, in addition to training data, making it unimaginable for the company to rebuild the technology.
This prospect is disturbing for 100 million people who use ChatGPT every week. This raises two questions that interest me as a law professor. First, can a federal court actually order the destruction of ChatGPT? And secondly, if he can, will he do it?
Destruction in court
The answer to the first query is yes. Under Copyrightcourts do indeed have the power to issue destruction orders.
To understand why, consider vinyl records. Their resurgent popularity attracts counterfeiters selling pirated records.
If a record company sues a counterfeiter for copyright infringement and wins, what happens to the counterfeiter’s stock? What happens to the master and stamp discs used to mass produce counterfeits and, above all, to the machines used to create these discs?
To answer these questions, copyright law gives courts the power to destroy infringing goods and the equipment used to supply them. From a legal perspective, a pirated vinyl record has no legal use. There can also be no valid reason for a counterfeiter to maintain a pirated master drive. Allowing them to maintain this stuff would only enable further violations of the law.
Therefore, in some cases, destruction is the only logical legal solution. And if a court finds ChatGPT to be like infringing goods or pirated hardware, it will possibly order its destruction. In its grievance, the Times argued that ChatGPT suits each analogies.
Copyright has never been used to destroy AI models, but OpenAI should take no solace in that. The law is becoming more open to the concept of attacking artificial intelligence.
Consider recent use by the Federal Trade Commission algorithmic distinction for instance. The FTC forced the firms equivalent to WeightWatchers to delete not only data collected unlawfully, but additionally algorithms and artificial intelligence models trained on such data.
Why ChatGPT will probably live to see one other day
It seems only a matter of time before copyright law is used to mandate the destruction of AI models and datasets. However, I do not think that can occur on this case. Instead, I see three more likely outcomes.
The first and simplest is that each parties can communicate. In case of successful settlement, which may be probablethe suit would be dismissed and no destruction ordered.
Second, the court may side with OpenAI by agreeing that ChatGPT is protected by the copyright doctrine “fair use” If OpenAI can argue that ChatGPT is transformative and that its service doesn’t replace content published by The New York Times, it may just win.
The third possibility is that OpenAI loses, but the law saves ChatGPT anyway. Courts can only order destruction if two conditions are met: first, the destruction must not prevent lawful activity, and second, it must be “the only cure” that might prevent violations.
This implies that OpenAI can save ChatGPT by proving that ChatGPT has legal and non-infringing uses or that its destruction will not be needed to stop further copyright infringement.
Both outcomes seem possible, but for the sake of argument lets say that the first destruction condition is met. The court could find that because of the articles included in the ChatGPT training data, any use infringed the Times’ copyright – an argument presented in various other lawsuits against firms engaged in generative artificial intelligence.
In such a scenario, the court would issue an injunction ordering OpenAI to stop infringing copyrights. Would OpenAI violate this order? Probably not. A single counterfeiter in a dodgy warehouse could attempt to get away with it, but that is less likely in the case of A $100 billion company.
Instead, it could attempt to train its AI models without using the Times articles, or it could develop other software safeguards to stop further problems. Given these possibilities, OpenAI would likely satisfy the second requirement and the court wouldn’t order the destruction of ChatGPT.
Given all of these obstacles, I imagine it is incredibly unlikely that any court would order OpenAI to destroy ChatGPT and its training data. But developers should know that courts do have the power to destroy illegal AI, and they appear to be increasingly willing to make use of it.