AI bots battling each other has led to the wrongful removal of educational content from one of YouTube's most respected mathematics channels. 3Blue1Brown, known for its excellent mathematical animations and explanations, recently had videos taken down due to false copyright claims generated by ChainPatrol.io.
The channel's creator, Grant Sanderson, posted on Twitter:
I learned yesterday the video I made in 2017 explaining how Bitcoin works was taken down, and my channel received a copyright strike (despite it being 100% my own content). The request seems to have been issued by a company chainpatrol, on behalf of Arbitrum, whose website says they "makes use of advanced LLM scanning" for "Brand Protection for Leading Web3 Companies" I could be wrong, but it sounds like there's a decent chance this means some bot managed to convince YouTube's bots that some re-upload of that video (of which there has been an incessant onslaught) was the original, and successfully issue the takedown and copyright strike request. It's naturally a little worrying that it should be possible to use these tools to issue fake takedown requests, considering that it only takes 3 to delete an entire channel.
YouTube's notoriously rotten-to-the-core AI-powered copyright system is a prime target of for tools can be manipulated to target legitimate content creators. Its copyright strike rule, which operates on a "guilty until proven innocent" basis is a constant source of annoyance for content creators. As one commenter noted in response to the incident, "There needs to be a reverse ban. For instance if you make three requests that turn out to be false, that company loses its ability to request copyright strikes."
ChainPatrol.io, the company whose system initiated the takedown, claims its "threat detection system makes use of advanced LLM scanning, image recognition, and proprietary models to detect brand impersonation and malicious actors targeting your organization." When it was called out for doing exactly the opposite, it acknowledged the error, tweeting:
Hello! This was a false positive in our systems at @ChainPatrol. We are retracting the takedown request, and will conduct a full post-mortem to ensure this does not happen again. We have been combatting a huge volume of fake YouTube videos that are attempting to steal user funds. Unfortunately, in our mission to protect users from scams, false positives (very) occasionally slip through. We are actively working to reduce how often this happens, because it's never our intent to flag legitimate videos. We're very sorry about this! Will keep you posted on the takedown retraction.v
For creators like Sanderson, who has spent years building an educational platform reaching millions, these automated systems pose an existential threat to their work. His fans didn't take kindly to ChainPatrol's weak sauce reply. Example tweets:
- Isn't that a crazy bad failure? The original was flagged?? It should be idiotically simple to validate by dates?
- F. U. Your system sucks and you suck.
- You say this because, it is suicide to claim 3b1b's video as yours. But if it was a smaller creator, you would have happily ignored. May the market show no mercy to you.
Previously:
• Facebook isn't free speech, it's 'algorithmic amplification optimized for outrage'
• New study reveals what we all know: YouTube's recommendation algorithms are terrible
• Facebook mistakes coding teacher for animal trader, bans him for life with no appeal
• HR software adds 'AI workers' as human employees
• Instagram's terrifying false positive nearly ruined a tech writer's life
