Nvidia is facing serious accusations for the massive collection of online videos for the purpose of training its artificial intelligence models. According to a report from 404 Media, a former anonymous employee of the company revealed that Nvidia has been using these videos not only for research, but also to improve commercial products such as Omniverse, its 3D world generator, and autonomous vehicle systems. The company allegedly instructed several employees to download videos, raising concerns about the legality and ethics of this practice.
Ming-Yu Liu, Nvidia’s Vice President of Research, claimed that these activities had “executive approval,” although controversy persists. The company seems to be expanding its role beyond AI hardware towards the development of foundational models using large volumes of data. Nvidia has indicated that it is building infrastructure capable of generating training data equivalent to a human lifetime of visual experience per day.
The controversy is part of a broader context of legal disputes over data collection for AI, affecting companies like OpenAI and Stability AI. As the debate over the legality of using data to train AI continues, Nvidia is working to solidify its position in a highly competitive market. Legislation surrounding these practices is still evolving, with proposals such as the AI Foundation Model Transparency Act in Congress.
Nvidia’s AI project, known as Cosmos, started in February 2024 and has collected 38.5 million URLs so far, with a significant percentage coming from movie videos. With a lack of clear regulations, Nvidia appears to be speeding up its development in a legally uncertain area.