Meta faces AI piracy suit after court rejects dismissal

Adult content producers Strike 3 Holdings and Counterlife Media can press copyright claims against Meta after U.S. District Judge Eumi K. Lee denied the company’s dismissal bid in California federal court.

The suit ties a familiar copyright fight to AI training. Strike 3, one of the most active U.S. copyright plaintiffs against BitTorrent users, accused Meta of using BitTorrent activity across corporate IP addresses to copy adult films. Counterlife joined the case with related claims.

Meta argued that IP address evidence cannot show corporate conduct by itself. The company said employees or visitors may have downloaded files for personal use, and it argued that some downloads predate its AI video training work.

Judge Lee let three claims move ahead: direct infringement, contributory infringement, and vicarious infringement. She rejected Meta’s argument that Strike 3 had to show the films entered a model training set. The alleged torrenting, she found, gave the plaintiffs enough for a direct copyright claim at the pleading stage.

That ruling matters for AI litigation because many rightsholders focus on model training. Lee’s order narrows the first fight here. The court can ask whether Meta copied protected works through BitTorrent before anyone proves that engineers fed those files into a model.

The order also gave weight to download patterns across 47 corporate IP addresses and seven hidden ranges. Strike 3 pointed to files with shared keywords and same-day downloads across Meta-controlled networks. The plaintiffs said those patterns looked coordinated.

Meta offered a coincidence theory. Lee found that explanation weak at this stage. The examples included files with “teen” in the title across unrelated categories, from cartoons to adult releases. The plaintiffs also cited multiple Meta IP ranges torrenting eight episodes of “Ted Lasso” out of sequence on one day.

Lee said those patterns support an inference that an algorithm searched and downloaded files by keyword. That finding does not prove Meta infringed. It gives Strike 3 and Counterlife enough room to seek records, depose witnesses, and test Meta’s account.

The court also addressed contributory infringement after the Supreme Court’s Cox Communications v. Sony decision raised the bar for plaintiffs. Meta argued that access to servers, data centers, IP addresses, computers, networks, and accounts cannot create liability by itself.

Lee accepted that limit but said Strike 3 alleged more. The plaintiffs accused Meta of creating tools and virtual private clouds that supported BitTorrent infringement. That allegation kept the contributory claim alive.

The vicarious infringement claim also stayed in the case. Lee found that Strike 3 tied the alleged copying to Meta’s financial interest in high-quality training data for commercial AI products.

Meta still has several defenses. The company says testimony from a related case shows its torrenting servers went live in 2024, not 2018. It also says much of the alleged activity occurred before Meta trained video models. Discovery will decide whether those timelines help Meta or undercut the plaintiffs’ story.

The dispute now moves into a fact-heavy phase. Meta must answer the complaint. The parties must attempt mediation by August, and the court scheduled a jury trial for February 2028.

For developers and AI teams, the order sends a plain signal: courts may treat data collection conduct as a separate copyright problem from model training. Teams that rely on scraped, torrented, or pooled media will need records that show who gathered files, how they gathered them, and why the company believed it could use them.