Copyright Law and the Inextricably Intertwined Futures of Journalism and Generative Artificial Intelligence

It’s increasingly evident that bright futures for both journalism and generative artificial intelligence (Gen AI) hinge on copyright law, licensing agreements, and high levels of cooperation between content creators and technology innovators. The New York Times Company’s December lawsuit against Microsoft and OpenAI highlights this reality and, as I previously described, the ramifications for a well-informed citizenry are profound.

Testimony during a January congressional hearing on “Oversight of A.I.: The Future of Journalism” reinforces this. Roger Lynch, CEO of magazine publisher Condé Nast, bluntly told the Senate Judiciary Committee’s subcommittee on privacy, technology, and the law that “Gen AI cannot replace journalism. It takes reporters with grit, integrity, ambition, and human creativity to develop the stories that allow free markets, free speech, and freedom itself to thrive.” In short, without journalism, the informational ecosystem that renders possible the benefits of Gen AI (and will likely generate billions of dollars for its investors) collapses. That’s because Gen AI’s large language models (LLMs) are typically trained on human-created, copyrighted content.

Via Reuters

This necessitates that common ground––financial and otherwise––be found between content creators who provide the informational backbone to a democratic society and technology innovators who seek, as OpenAI grandiosely proclaims, “to ensure that artificial general intelligence benefits all of humanity.” More provocatively, the muckrakers and buckrakers now must make merry together.

As Lynch sees it, however, such a symbiotic relationship is nonexistent today. Instead, “Gen AI companies copy and display our content without permission or compensation in order to build massive commercial businesses that directly compete with us.” Furthermore, “Gen AI tools hallucinate and generate misstatements that are sometimes attributed to real publications like ours, damaging our brands.”

This echoes allegations in The New York Times Company’s complaint that Microsoft and OpenAI are “using the The Times’s content without payment to create products that substitute for The Times and steal audiences away from it.” Additionally, The Times contends the defendants’ Gen AI tools harm the newspaper “by misattributing content to The Times that it did not, in fact, publish. In AI parlance, this is called a ‘hallucination.’ In plain English, it’s misinformation.” The Times’s readers, in turn, are harmed because false attributions lead “them to incorrectly believe that the information provided has been vetted and published by The Times.”

Lynch, a former CEO of audio streamer Pandora and founder of live-streaming service Sling TV, points the way forward on this critical issue: AI companies must obtain licenses for the copyrighted content on which their business model depends. Such a payment system allowed prior disruptive media innovations like Internet Protocol TV and Pandora to flourish, Lynch asserts. He testified that such

successful new businesses were built on a foundation of licensing content rights. Licensing allowed distributors to work together with content creators to innovate new and better consumer experiences and generate profits that were reinvested in great content. . . . congressional intervention is needed to make clear that Gen AI companies must also seek licenses to utilize publisher content for use with Gen AI.

Of course, one hopes congressional action isn’t necessary and that mutually advantageous licensing agreements like the one between the Associated Press news service and OpenAI are freely negotiated.

It’s more than just the use of copyrighted magazine and newspaper content in training LLMs, however, that raises significant problems. Curtis LeGeyt, president and CEO of the National Association of Broadcasters, testified that:

the use of broadcasters’ news content in generative AI models, without authorization or compensation, risks further diminishing reinvestment in local news. Broadcasters have already seen numerous examples where content created by broadcast journalists has been ingested and regurgitated by AI bots, with little or no attribution. Not only are broadcasters losing out on compensation for their own work product, but this unauthorized usage actually increases costs for local stations due to additional vetting of stories and footage and the costs associated with protecting broadcast content.

DeGeyt also asserted that the “lack of attribution and sourcing in AI-generated outputs” harms the public by making “it difficult to identify and distinguish legitimate, copyrighted broadcast content, from the unvetted and potentially inaccurate content being generated by AI.” This all comes, he pointed out, when “local news production is increasingly costly.”

DeGyet, however, was upbeat about potential benefits of AI when journalists themselves deploy it. For example, he explained that AI can translate stories into different “languages to better serve . . . diverse audience[s]” and can “convert broadcast scripts––written by the station’s local journalists––into digital stories that are also accessible on a local station’s website.” Additionally, Gen AI can provide “first drafts of content for human review.”

In sum, old-school journalists and new-tech innovators should and must help each other and, in turn, help the public and democracy through copyright cooperation, not uncompensated exploitation.