In the evolving landscape of artificial intelligence, new search technologies from leaders like OpenAI and Microsoft are reshaping how news companies interact with AI entities. These AI-driven tools, particularly those that use news content to answer real-time queries, are prompting a shift in negotiations between news publishers and technology firms. Historically, discussions have revolved around data provision for the training of large language models. However, current conversations are focusing on more specific use cases, potentially giving publishers increased bargaining power.
Large language models are trained on extensive datasets compiled from a myriad of sources. To enhance their accuracy, especially for queries related to current events, these models are now incorporating a technique known as Retrieval Augmented Generation (RAG). RAG not only helps in refining the responses of these models but also aims to mitigate the issue of “hallucinations”—the generation of incorrect or fabricated answers.
The introduction of OpenAI’s SearchGPT and Microsoft’s Bing generative search engine highlights a burgeoning trend where AI companies integrate RAG to improve the relevancy and credibility of their search results. OpenAI, for instance, is piloting SearchGPT with media outlets like The Atlantic and News Corp, ensuring that responses provide transparent attribution to news content sources directly within the search interface.
Despite these advances, it remains uncertain whether AI-powered search engines will match the revenue generation capabilities of traditional search engines like Google, which have benefitted publishers through traffic-driven income. OpenAI is experimenting with licensing fees rather than revenue sharing in its arrangements with news publishers, while startups like Tollbit explore marketplace models where profits are distributed based on user engagement.
This tension between AI firms and publishers is underscored by legal debates over copyright issues. Publishers argue that their content, even if publicly accessible, is protected under copyright law, leading to high-profile lawsuits like that of The New York Times against OpenAI and Microsoft. These legal battles seek to clarify the extent to which AI can utilize copyrighted materials without direct authorization.
In response to the legal uncertainties and the evolving needs of AI technologies, both small and large tech companies are navigating the complexities of media partnerships differently. While OpenAI actively pursues agreements with newsrooms, others like Anthropic remain noncommittal about forming publisher alliances. Meta, with its history of fluctuating commitment to media partnerships, is currently deliberating its approach, considering whether engaging with news providers will be essential for the accuracy of its AI products.
Perplexity, on the other hand, is testing a revenue-sharing model with various publishers, including Time and Der Spiegel, as part of a pilot program aimed at balancing content utilization with equitable compensation. However, despite these efforts, resistance from media companies persists, with entities like Forbes and Condé Nast actively challenging unauthorized use of their content.
As the media and AI industries continue to intersect, the outcomes of these negotiations and legal challenges will likely shape not just the future of news consumption but also the broader dynamics of content creation and distribution in the digital age. For more detailed insights, read the full report on Axios: AI’s search quake shakes media landscape.