Publishers Demand Billions From Chatbots For Their Data

Publishers Seek Orders of Magnitude Higher

Barry Diller, chairman and senior executive of AIC, fired the first shot on behalf of publishers against artificial intelligence platforms during an April interview with Semafor. In his remarks, he suggested that these companies should be sued for using publishers’ data to train their AI models.

Now, Diller’s company, IAC, along with a select group of influential publishers, is on the verge of forming a coalition to pursue legal action and advocate for legislative measures. Notably, this coalition includes industry giants such as The New York Times, News Corp., and Axel Springer.

 

In an interview at his Manhattan office, Joey Levin, the current CEO of IAC, who is playing a significant role in the coalition, expressed concerns about the potential impact of generative AI on the media. He stated that the AI takeover of the media could have far-reaching consequences beyond “the typical sci-fi fears.”

As AI tools that aim to streamline writing become more prevalent, publishers are experimenting with them. However, executives are also worried about various threats, including potential impacts on revenue and the very essence of online authority.

One of the immediate concerns is the possibility of Google shifting from sending traffic to web pages to solely responding to user queries with chatbots. For instance, this could lead to a Food & Wine review being replaced with a simple text recommendation for a bottle of Malbec, without proper attribution – a scenario that worries Levin. Google has mentioned definitive plans to implement AI on their search pages.

He emphasized that these large language models, or generative AI systems, are designed not only to find valuable information but also to potentially “steal” the best content from the internet.

Previous Settlements from the Early Social Media Days of Web 2.0

While publishers and tech executives aim to avoid repeating past conflicts and controversies, a resolution regarding compensation for AI training data is being sought. Tech companies hope to mollify publishers with payouts in the millions, similar to what the Facebook News Initiative provided previously. However, publishers are pushing for much larger sums, asserting that their contributions to training AI language models should be proportionally rewarded – potentially reaching billions of dollars across the industry.

As the two sides attempt to find common ground, the future of AI in the media industry remains uncertain, and the possibility of legal battles looms large. Tech executives argue that they have yet to establish a profitable business model for AI, while publishers insist on fair compensation for their invaluable data inputs. This ongoing dispute is likely to escalate unless a mutually agreeable solution is reached.

During discussions involving prominent figures from both sides of the argument, it is becoming evident that a potential resolution is emerging: AI companies may be required to compensate publishers for their training data. However, there is one significant obstacle.

Similarities and Differences With Past Settlements

Tech companies seem to be aiming to appease publishers by offering payouts in the eight-figure range, similar to the approach taken by the Facebook News Initiative, which provided payments annually from 2019 to 2022. These fees were reported to exceed $20 million for the Times, $15 million for the Washington Post, and $10 million for the Wall Street Journal. The ongoing tension of content creators who realize what they were giving away on the internet was undervalued. Reddit communities recently in light of the AI changes and the ongoing writers’ strike are other examples.

Yet, publishers firmly believe that the sums offered should be substantially larger this time. Their argument is based on the fact that the breakthrough language models rely heavily on the publishers’ valuable inputs, and consequently, the compensation should be proportionate, potentially reaching billions of dollars across the entire industry.

Neither side has disclosed specific figures, nor have they openly discussed the coalition they are forming. However, publishers, led by Barry Diller himself, have made it clear that if the proposed settlement falls short of their expectations, they may resort to legal action. They are closely observing an ongoing Delaware lawsuit involving an AI company’s copying of legal texts from Westlaw, as the outcome could have implications on copyright law related to AI training data and outputs.

Such payments on the scale envisioned by publishers could bring about a significant transformation for companies like Google, which have thrived on high-margin businesses largely due to their avoidance of content payments, unlike media companies such as Netflix and Comcast.

Ultimately, the resolution of this high-stakes conflict depends on whether the publishers are willing to reconsider their demands or if the tech companies will adjust their fundamental notion of what it means to be a platform. Until then, the tension is likely to escalate.