Recently, an internal electronic spreadsheet obtained by 404Media revealed shocking secrets about Runway's Gen-3Alpha video generator, a highly anticipated AI video generation tool. It turns out that this tool was trained on a vast amount of pirated content and unauthorized YouTube videos!
Gen-3Alpha garnered widespread attention at its release last month, with many media outlets praising its generated visuals for being almost lifelike. At the time, Runway only stated that Gen-3Alpha was "trained jointly on videos and images" without detailing the data sources. Now, this spreadsheet unveils more insider information—it contains numerous video links from major YouTube channels like Disney, Netflix, Sony, and some links from well-known piracy websites.
Although 404Media could not confirm whether all listed assets were used for Gen-3Alpha's training, based on the available information, this is highly likely. This incident again exposes the disrespect for copyright by AI companies when generating content, especially when using a large amount of unauthorized materials to train AI models, which has become a persistent issue.
Reports indicate that Runway even used proxy servers to bypass YouTube's blocks to download all these videos. Runway raised $141 million last year, with investors including Google (YouTube's parent company), Salesforce, and chip maker NVIDIA, valuing the company at $1.5 billion. However, this reliance on pirated content could expose the company to severe legal challenges.
In addition to Runway, OpenAI has also come under criticism for using unauthorized video data. Earlier this year, OpenAI's Chief Technology Officer stated in an interview that it was uncertain whether the company's newly developed video generator Sora used videos from YouTube, Instagram, or Facebook. Subsequently, The New York Times revealed that OpenAI had improperly used YouTube videos in training its AI chatbot.
YouTube CEO Neal Mohan has also warned AI companies that using YouTube videos to train AI models would be a "clear violation" of the video platform's terms of use.
Copyright issues are becoming a significant obstacle to the development of generative AI technology, especially for AI models that can generate complete videos. Legislators are re-examining the legal principle of "fair use" to address the challenges posed by this technology.
**Key Points:**
📊 **Runway's Video Generator Exposed for Using Pirated Content** — Runway's Gen-3Alpha video generator was trained on unauthorized YouTube videos and pirated content, sparking copyright controversies.
🛡️ **AI Companies Repeatedly Violating Copyright Laws** — This issue is not limited to Runway; companies like OpenAI also face similar problems for using unauthorized data, making copyright disputes a bottleneck for AI development.
📜 **Legislators Re-examining Copyright Regulations** — As AI technology advances, related laws and copyright policies are also being updated to adapt to new technological challenges.