We missed this earlier: Social media conglomerate Meta is involved in revenue-sharing agreements with hosts of its Llama Artificial Intelligence (AI) models, revealed court filings related to an ongoing copyright lawsuit, Kadrey vs. Meta.
While the case itself deals with Meta’s alleged usage of a vast quantity of data from pirated e-books to train Llama, proceedings before the United States District Court for the Northern District of California on March 21 saw plaintiffs’ attorneys noting that the company “shares a percentage of the revenue” that the AI model hosts generate from Llama users.
The court redacts the names of the AI model hosts in question with whom the company might have such arrangements in place. However, a TechCrunch report lists companies like AWS, Nvidia, Databricks, Groq, Dell, Azure, Google Cloud, and Snowflake as hosting partners for Llama which Meta has purportedly mentioned in several of its blog posts.
Meta Does A U-turn on AI model sharing
Interestingly, this comes after Meta CEO Mark Zuckerberg said back in July 2024 that the company was not keen on the prospect of “selling access to AI models”, while pushing for wider adoption of open sourcing as a practice.
But he did have plans in mind for monetising aspects of the service. As per the report, during Meta’s earnings call in April 2024, the former floated ideas such as monetisation through business messaging services and advertisements in “AI interactions”.
“[I]f you’re someone like Microsoft or Amazon or Google and you’re going to basically be reselling these services, that’s something that we think we should get some portion of the revenue for,” Zuckerberg was quoted saying.
Why it Matters
Apart from revenue-sharing agreements for Llama, an AI model allegedly trained through pirated e-books among others, the court document revealed something else that is significantly worrisome.
“Meta’s engineers were well aware of the legal risks posed by pirating data through P2P sharing, but they decided to do it anyway, apparently with the approval of Meta’s in-house counsel,” the release stated.
Advertisements
It also described messages exchanged between several Meta engineers regarding the legitimacy of ‘seeding’ (uploading) the concerned files, as it can land the content outside, potentially having legal consequences. One of the examples it provided in this regard is the struggle they faced with directly downloading shadow library LibGen, eventually deciding to Torrent the entire SciMag section of LibGen that contained “approximately 81 million scientific articles.”
While Meta’s actions, such as the back-and-forth on monetisation and revenue sharing with partners for Llama, are largely discussable here, what is also noteworthy is the copyright argument against the company and the alleged ways in which it has tried to bypass ethical responsibilities by opting for dubious strategies like torrenting. This raises questions as to whether stricter regulation is required in the AI model training/development space.
Also Read:
Support our journalism: