A video-3D foundation model was trained on YouTube videos, according to its developers, prompting accusations of violations of YouTube's terms of service.

Viggle's CEO, Hang Chu, inadvertently confirmed in a TechCrunch interview that the company's AI video models had been trained on publicly available data, including YouTube videos. 

YouTube CEO Neal Mohan had previously stated that using videos from the platform to train AI models without permission is a "clear violation" of YouTube's rules. This is seen to have put Viggle's practices in a potentially problematic position.

The use of publicly available content for AI training is a common practice in the technology industry, but it remains a contentious issue that few companies openly discuss. Viggle's case highlights the ongoing debate about the ethical and legal implications of using such data for AI development.

Following the interview that revealed their use of YouTube videos, a Viggle spokesperson clarified that the company is careful to comply with terms of service, avoiding unauthorised downloads and other violations. However, the specifics of how they ensure compliance were not detailed.

This controversy is far from unique to Viggle, and reflects a broader challenge in the AI industry as AI technologies advance and questions about data sourcing, copyright, and fair use continue to emerge.