Runway, a startup backed by Google, is facing backlash from creators after a report by 404 Media alleged that the company copied data from thousands of YouTube videos. The report claims that a former employee leaked a company spreadsheet showing Runway’s plans to use data from YouTube channels of media and entertainment companies, as well as individual YouTubers, to inform its AI video creation model called “Jupiter.” This revelation has rubbed many creators and critics the wrong way, as it suggests that Runway may have copied, downloaded, or analyzed YouTube videos without permission.
The leaked spreadsheet, which has been published online, includes over 3,900 individual YouTube channels and a column with hashtags representing different content. It also contains a tab labeled “high_camera_movement” with more than 177 distinct YouTube accounts. While it is unclear if all the videos in the spreadsheet were used to train Runway’s Gen-3 Alpha model, the existence of the spreadsheet itself raises ethical concerns among creators and critics of generative AI.
Influential tech reviewer Marques Brownlee, who has been critical of others training AI on his videos in the past, expressed his displeasure on Twitter. He posted a link to the report and commented, “Well well well. Runway AI video generator was trained on YouTube videos without permission, including 1600+ MKBHD videos.” However, it is worth noting that Brownlee has also shown excitement for AI video technology like OpenAI’s Sora in a previous video.
Ed Newton-Rex, founder and CEO of ethical AI certification startup Fairly Trained, highlighted the notable names included in the alleged Runway spreadsheet, such as musician Taylor Swift and filmmaker Wes Anderson. Other creators, like YouTuber Omni, called the spreadsheet “INSANE” and accused Runway of theft. Even AI filmmakers who have used Runway’s tools in the past, like Dustin Hollywood, criticized the company for what they perceive as data theft.
It is important to note that multiple companies, including Apple, Nvidia, and the AI startup Anthropic, have also been accused of using copyrighted videos without permission or payment for training their AI models. While scraping and training may be controversial, it is supported by the precedent set by Google in web scraping for search indexing. However, Runway and other AI companies are facing lawsuits from creators who claim that their data was used without permission or compensation. In the court of public opinion, Runway has undoubtedly suffered a blow due to these allegations.
In conclusion, the controversy surrounding Runway’s alleged data copying from YouTube videos has ignited a debate about the ethics of using copyrighted content for AI training. Creators and critics have expressed their concerns and called out the company for potential theft. While the legality of scraping and training is still up for debate, the court of public opinion has already cast a negative light on Runway’s actions. The outcome of any legal action taken against the company will determine the future implications for AI companies and their use of copyrighted material.