Revenue sharing agreements between Meta and companies hosting its Llama AI models have been uncovered through an unredacted court filing in the Kadrey v. Meta copyright lawsuit1. While specific hosting partners are not named, Meta has previously listed collaborators including AWS, Nvidia, Databricks, Azure, Google Cloud, and Snowflake2. Under these arrangements, Meta "shares a percentage of the revenue" generated by users leveraging the Llama models through these hosting services13. This revelation suggests a shift in Meta's AI monetization strategy, potentially helping to offset the company's substantial investments in AI infrastructure, with planned capital expenditures of $60-80 billion in 2025 largely focused on data centers and AI development teams4.

Mark Zuckerberg's previous assertion that "selling access" to Meta's openly available Llama AI models "isn't [Meta's] business model" has been called into question by the recent court filing1. This revelation of revenue-sharing agreements directly contradicts Zuckerberg's earlier stance, highlighting a potential shift in Meta's approach to monetizing its AI capabilities. The discrepancy between the CEO's public statements and the company's actual practices raises concerns about transparency and could potentially impact Meta's credibility in the AI community23.

The revelation of Meta's revenue-sharing agreements has sparked debate over the company's commitment to open-source principles in AI development. While developers can still download and run Llama models independently, the financial arrangements with hosting partners offer convenience and additional services, potentially blurring the line between open-source and commercial interests12. This development highlights the complex balance between fostering innovation through open-source initiatives and pursuing commercial viability in the AI industry3. As a result, Meta's approach could set a precedent for how tech companies navigate similar challenges in the future, potentially influencing the broader landscape of AI development and distribution.

The Kadrey v. Meta lawsuit has unexpectedly shed light on Meta's AI business practices, particularly regarding its Llama models. This copyright infringement case alleges that Meta trained its Llama models on hundreds of terabytes of pirated ebooks1. The unredacted court filing in this lawsuit revealed Meta's revenue-sharing agreements with Llama AI model hosts, contradicting previous public statements about the company's AI monetization strategy12.

This legal battle not only exposes potential copyright issues in AI training data but also highlights the complex interplay between open-source initiatives and commercial interests in the AI industry. As the case progresses, it may set important precedents for how tech companies handle data sourcing for AI training and how they balance open-source commitments with revenue generation strategies34.