Bloomberg reports frontier models used distillation despite copyright concerns

Bloomberg reported that leading AI laboratories allegedly employed model distillation techniques during early training phases despite internal awareness of potential copyright infringement risks. The report claims companies prioritized rapid capability scaling over intellectual property compliance, suggesting current market leaders may have built foundational advantages through contested data practices. Critics argue this alleged conduct undermines fair use legal defenses by demonstrating bad faith regarding creator rights. Industry representatives maintain that distillation constitutes standard technical optimization rather than willful infringement. Legal experts warn that if courts accept evidence of deliberate IP evasion during training, it could invalidate transformative use arguments central to current AI copyright litigation. The allegations emerge as multiple lawsuits against major AI developers proceed through discovery, potentially exposing internal communications about data sourcing strategies. Bloomberg cited unnamed sources familiar with company operations in making these claims, which remain unadjudicated allegations subject to ongoing legal proceedings.