Courts Force Tech To Name AI Approvers

A magistrate ordered Meta to hand over server command histories and to identify employees who approved or ran BitTorrent downloads of shadow‑library datasets (Anna's Archive, LibGen) used to train Llama models. The ruling treats decisions about acquiring training data — and the technical acts that seed torrents — as discoverable and potentially legally actionable. That pressure could reshape corporate data‑acquisition workflows and internal legal sign‑offs for AI training. — If courts can force disclosure of who approved dataset procurement, companies will face stronger legal incentives to document, limit, or change how they gather training data for AI, affecting transparency, liability, and model development practices.