Meta has refuted allegations that it utilized adult videos to train its AI models. The company clarified that while some employees may have downloaded a limited number of videos for personal reasons, this was not for AI development purposes. The lawsuit, filed by Strike 3 Holdings, a producer of adult films, claims that Meta employed its videos to train an undisclosed AI tool named Movie Gen and seeks $350 million in damages. In response, Meta requested that a US court dismiss the lawsuit, asserting that there is no evidence supporting the claim that it trained AI on adult content.
The company noted that only approximately 22 adult videos were downloaded annually from its network, suggesting this activity might stem from individual users. Furthermore, Meta emphasized its policy against incorporating adult content in its AI training. A spokesperson stated, “We don’t want this kind of content and work to avoid it.” This legal issue arises amid reports of Meta’s AI chatbots generating unsafe or inaccurate responses, prompting the company to implement more stringent regulations for its AI systems. Earlier this year, a US court ruled in favor of Meta in a different case, determining that the use of books for training its Llama AI models constituted fair use under US law.