Reliable sources have revealed that 404 Media has successfully acquired a vast internal training data table. The content revealed by this table is shocking. It turns out that Runway, a company worth billions of dollars, has secretly used a large number of videos from well-known YouTube creators and brands, numbering in the thousands, as well as unauthorized movie content, in the training process of its highly praised AI video generation tool. The insider who leaked this significant news has also provided detailed information about these YouTube channels, allowing people to better understand the issues with the training data sources of the tool.
In June of this year, the technology news site Techcrunch inquired about the source of training data for Runway's Gen-3 to Runway's co-founder Anastasis Germanidis. However, Germanidis did not disclose any specific details in his response. He simply stated to Techcrunch, "We have a dedicated internal research team that strictly oversees all training processes. We use carefully selected internal datasets for training our models." However, given the current exposure, his response seems to be significantly inconsistent with the actual situation, raising questions about the company's compliance with data usage regulations.