Google Pays Publishers for AI Training Data as Hollywood Licensing Talks Stall

Google has launched a pilot program that pays publishers for their content used in AI-powered search features, signaling a shift in how the company acquires training data. According to The Information, around 100 publishers have joined the program, which began less than a year ago. Payments are tied to how much a publisher's content contributes to AI Overviews, AI Mode in Search, and the Gemini chatbot. One early participant earned over $1 million in a year, while another that joined a few months ago has made between $50,000 and $60,000, as reported by The Verge.

The pilot comes as Google faces regulatory pressure over AI's impact on web traffic. In June, the UK ruled that Google must allow publishers to opt out of AI search features, and the European Union opened an investigation and ordered changes to its search engine. The program is an attempt to address criticism that Google's AI features cannibalize publisher traffic without compensation.

Meanwhile, Google's separate effort to license Hollywood studios' libraries for AI training has hit a wall. As reported by The Verge citing The Los Angeles Times, Google has pitched deals to Disney, Warner Bros. Discovery, and Universal, offering hundreds of millions of dollars for access to copyrighted characters and footage. For example, Google is willing to pay around $40 million for the rights to generate outputs featuring a single Disney/Pixar character, with potential payouts swelling into the billions for broader libraries. However, no major studio has agreed to a deal. The studios face risks: licensing content to Google could devalue their own IP and alienate audiences, especially if AI-generated content floods platforms like YouTube.

The contrast between the two strategies is instructive. Publishers, whose business models depend on search traffic, have been eager to join the pilot. Studios, with more leverage and alternative revenue streams, are holding out. For Google, the stakes are high. Training AI models on high-quality, copyrighted material is seen as a competitive advantage over rivals like OpenAI and Meta. But the Hollywood standoff suggests that even with deep pockets, Google cannot simply buy its way into every content library.

A separate case involving Spirit Airlines' bankruptcy underscores the broader tensions around AI data acquisition. Ars Technica reports that Google won an auction to buy a huge dataset from Spirit, but a startup called Springshot has objected, arguing that Spirit does not own the data. Springshot's platform powered Spirit's operations for three years, and the company claims the sale includes its intellectual property. The bankruptcy court has been urged to pause the sale until a forensic audit establishes ownership. As Springshot's founder put it: "Bankruptcy cannot become the new land grab for AI."

For engineers and founders building with AI, these developments highlight a critical challenge: the provenance and legality of training data. Google's publisher pilot offers a transparent, contractual model. The Hollywood talks show that even with billions, some content remains off-limits. And the Spirit case is a warning that data ownership disputes can erupt in unexpected places — including bankruptcy proceedings. As AI companies scramble for training material, the rules of the road are still being written, one deal and one lawsuit at a time.