The limits of copyright law are starting to be tested by the explosion of artificial intelligence systems that access vast amounts of content from the internet.
Over the past year, authors and a top photo agency have filed lawsuits, claiming that AI systems—which can generate language that resembles that of a human being and power chatbots—have been trained illegally using their intellectual property.
The journalism industry has now joined them in the limelight. The New York Times is the first major American news organization to fight Microsoft and OpenAI over the use of artificial intelligence through a lawsuit filed on Wednesday, alleging copyright infringement.
Claiming that Microsoft’s Bing Chat and OpenAI’s ChatGPT can generate content almost exactly like Times articles, the lawsuit says that these two platforms are able to “free-ride on The Times’s massive investment in its journalism by using it to build substitutive products without permission or payment.”
Microsoft and OpenAI are not in a position to answer in court. However, following the lawsuit’s filing, those businesses disclosed that they were in talks about utilizing content with other news outlets; in OpenAI’s case, they had even started to execute agreements.
In the absence of such agreements, the boundaries might be established by the courts, which would have serious consequences. The development of generative AI technologies, which can produce text, images, and other media on their own, as well as the business plans of the companies that carry out that work, depend heavily on data.
One of the major factors influencing the generative AI market, according to analyst Fred Havemeyer of Macquarie Financial Research, will be copyright.
In intellectual property law, one important factor to take into account is the “fair use” doctrine, which permits producers to expand upon copyrighted material. Defendants in copyright lawsuits must demonstrate, among other things, that they significantly altered the content and are not competing in the same market as the original creator’s work.
For example, a review that incorporates quotes from a book could be deemed fair use since it expands on the original content to produce a new piece of writing. On the other hand, selling long chunks from the book could go against the doctrine.
Regarding how those criteria relate to AI tools, courts have not commented.
Intellectual property attorney Ryan Abbott of Brown Neri Smith & Khan stated, “There isn’t a clear answer as to whether or not that is copyright infringement in the United States.” “There are numerous lawsuits pending in the interim that could be worth billions of dollars.”
It can take some time before the industry receives conclusive responses.
The cases raising these issues are still in the early phases of the legal process. It might take years before a Federal District Court makes a decision in this case if they don’t result in settlements, which is what happens with most lawsuits. These verdicts would almost certainly be challenged, and since appellate decisions differ by circuit, the case might end up before the US Supreme Court.
According to Abbott, getting there might take ten years or more. “In the market we are currently in, ten years is an eternity,” he remarked.
The Times stated in its lawsuit that it had discussed terms for settling the dispute, potentially involving a license, with Microsoft and OpenAI. OpenAI and the Associated Press have recently established data license deals. Axel Springer is the German owner of Politico and Business Insider.
Trials could provide important information regarding what and how copyrighted data AI developers can utilize. However, it could also only be used as leverage by the plaintiff in a settlement negotiations to obtain a better licensing agreement.
Professor Jane Ginsburg of Columbia Law School commented on the lawsuit by The Times, saying, “Ultimately, whether or not this lawsuit ends up shaping copyright law will be determined by whether the suit is really about the future of fair use and copyright, or whether it’s a salvo in a negotiation.”
The evolving legal environment may have an impact on the highly capitalized but still-developing AI sector.
Following ChatGPT’s public launch, which went viral last year, other AI startups have received a deluge of venture cash. OpenAI is valued at around $80 billion according to a stock plan that is being considered; Microsoft has invested $13 billion in the company and has integrated its technology into its own products. However, investors have expressed concern over the use of intellectual property to train models, according to Mr. Havemeyer.
In the field of AI, competition could ultimately come down to data haves and have-nots.
Businesses that have acquired their own data, like Google and Meta, or that possess the rights to vast amounts of data, like Adobe and Bloomberg, have begun creating their own artificial intelligence systems. Havemeyer pointed out that a well-known business like Microsoft was well-suited to negotiate data licensing terms and handle legal issues. However, it can be more difficult for startups with limited funding to get the data they require to compete.
Havemeyer stated, “Data is the beginning and the end of generative AI.”
