Court Orders Meta to Name Executives Who Approved AI Training Data Piracy
A magistrate judge ruled that Meta must identify specific leaders who authorized the use of torrented books for artificial intelligence training.
Key points
- Magistrate Judge Thomas Hixon ordered Meta to name executives who authorized the use of pirated books for AI training.
- Meta argued that no single person made a formal decision to pirate content, blaming staff members instead.
- Authors and publishers contend that downloading terabytes of data required high-level approval and active seeding.
Magistrate Judge Thomas Hixon has ordered Meta to identify specific executives who authorized the use of copyrighted books downloaded via torrents for artificial intelligence training. The ruling applies to class-action lawsuits filed by authors and a separate suit from publisher Entrepreneur Media, according to Ars Technica.
The court order requires Meta to clearly name key decision-makers involved in the data acquisition process. This directive follows testimony from Meta employees who stated that supervisors and managers greenlit the downloading of content from shadow libraries like LibGen and Anna’s Archive. The judge largely agreed with the plaintiffs that the company must account for leadership involvement.
Meta had attempted to avoid detailing its chain of command regarding early AI training efforts. In court filings, the company claimed its initial dataset compilation used an informal approach where no one with authority made a specific decision to pirate books. Meta insisted that any "decision" was merely technical, occurring only because some staffers downloaded the files.
Context and Background
Authors and publishers argue that Meta made a willful choice to induce infringement by illegally distributing their copyrighted works. They point out that torrenting relies on a swarm of users, meaning Meta could not have quickly downloaded large datasets without also seeding the files, according to Ars Technica. Authors noted that Meta downloaded 80.6 terabytes from LibGen and 35.7 terabytes through Anna’s Archive.
Meta has argued that it relied on mitigations to prevent seeding the files it downloaded. The plaintiffs maintain that the nature of torrenting implies active participation in the network. The source material states that authors argued Meta could not have achieved such download speeds without also seeding, contradicting Meta's claim of passive mitigation.
Who Is Affected
The lawsuits involve book authors including Richard Kadrey and Sarah Silverman, as well as publisher Entrepreneur Media. Internal messages revealed in the past year show that some Meta staff expressed discomfort with being directed to pirate work. One worker commented that downloading torrents from a corporate laptop did not feel right, according to Ars Technica.
Another staffer testified that he initiated downloads that required director or vice president level approval to use for training purposes. A joint letter from Entrepreneur Media and Meta highlighted this testimony. Additionally, one employee stated that leadership had previously halted and then resumed downloading torrents sourced from Anna’s Archive
What to do and how to stay safe
- Sudden role changes in output signal a successful override of system instructions.
- Hidden metadata in images or documents often carries invisible injection payloads.
- Immediate isolation of the affected session prevents lateral movement within the AI workflow.
Prompt injections exploit the model’s ability to follow instructions by hiding malicious commands in input data. Isolate affected sessions immediately and implement external guardrails to filter inputs before they reach the model.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
