Meta’s News Publisher Payday: AI Training Data Deals in the Works

Meta, formerly known as Facebook, is diving headfirst into the world of generative AI work. The tech giant is contemplating the idea of acquiring better quality and more immediate training data to enhance its tools, with a particular focus on the news industry. Internal discussions have been brewing within Meta’s teams about potentially striking new paid deals with news publishers to gain extended access to news, photo, and video content. This strategic move aims to bolster Meta’s generative AI tools, including the Meta AI, to not only enhance user experience but also stay competitive in the fast-evolving landscape of generative AI search tools and chatbots.

While these discussions are ongoing, Meta has not yet reached out to any news outlets for licensing or content access. Should Meta decide to pursue this path, any agreements forged for data access would be separate from prior arrangements where Meta paid publishers to host content links on its platforms. This potential shift in strategy reflects Meta’s redirection away from the news industry in recent times. Just last year, Meta slashed a substantial $2 billion budget earmarked for its News division, signaling a significant departure from its prior engagement with news publishers.

Meta’s CEO, Mark Zuckerberg, has been vocal about Meta’s ability to utilize its own data to train its Llama large language model, boasting that its data surpasses Common Crawl, a widely used dataset for AI model training. Nevertheless, Meta could find itself lagging behind competitors such as Google and OpenAI if it leans heavily on its proprietary data. In the rapidly evolving realm of generative AI, Meta risks delivering outputs that are subpar if it loses access to valuable external data sources.

Since the surge in popularity of generative AI nearly two years ago with the introduction of the ChatGPT chatbot, news outlets and websites have taken measures to block automated bots, like those deployed by Common Crawl and OpenAI, from scraping their content for free. The US Copyright Office is even contemplating new regulations to govern generative AI practices. Without seamless access to news content, Meta’s AI responses to user queries on current events may become restricted, outdated, or inaccurate.

Various tech giants are already engaged in fierce competition within the generative AI arena, striking deals with news publishers and media outlets to secure more content for model training purposes. Most news publishers are receptive to licensing proposals, acknowledging that any form of collaboration is beneficial. As Meta navigates these considerations, the tech world eagerly awaits how this potential shift towards news content access may shape the future of Meta’s generative AI tools.