📰 New article from TorrentFreak
Rightsholders Can’t Use OpenAI and Anthropic to Dismantle Meta’s Seeding Defense
### The “It’s Just How BitTorrent Works” Defense
Meta is currently in a legal tug-of-war over how it trained its Llama AI models, and the argument is getting surprisingly technical. Authors and publishers are accusing Meta of using pirated “shadow libraries” to feed its AI, and because Meta used BitTorrent to get the data, they ended up “seeding” (uploading) those pirated books to other users in the process.
Meta’s defense? They claim that uploading is an “inherent characteristic” of the BitTorrent protocol—basically saying, “Look, if you want the data in bulk, you have to share it. It’s just how the tech works.”
In a clever move, plaintiffs tried to call Meta’s bluff by subpoenaing AI rivals OpenAI and Anthropic. The logic was simple: if OpenAI and Anthropic used torrents but configured their software not to upload, then Meta’s “it’s unavoidable” excuse falls apart.
However, the court just shut that plan down. A judge ruled that poking around OpenAI’s logs wouldn’t prove anything about Meta’s specific settings. Instead of playing detective with Meta’s competitors, the court says the plaintiffs should look under Meta’s own hood.
The takeaway: Meta has been ordered to hand over its own server command histories. This will show exactly how their torrent clients were configured—and whether they intentionally chose to seed pirated content or if it really was “just the protocol.”