News: 1736212263

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

3Blue1Brown copyright takedown blunder by AI biz blamed on human error

(2025/01/07)


The errant copyright takedown of a popular video explaining Bitcoin, from Grant Sanderson's smash-hit YouTube channel 3Blue1Brown, has been reversed, with human error blamed.

The brand protection company responsible for the unwarranted takedown, ChainPatrol.io, also acknowledged that the legal demand was a mistake, and has been in communication with Sanderson about its efforts to return the video to YouTube.

Sanderson – who has 6.8 million subscribers on YouTube, and whose [1]in-depth videos mainly on mathematics and science have been viewed more than 600 million times – raised the alarm in a social media [2]post on Monday.

[3]

He said he had just learned that YouTube had removed a 2017-era video he had made and applied [4]a copyright strike – three of which will get an account terminated – despite the fact that he had only used his own content in his video.

[5]

[6]

"The request seems to have been issued by the company ChainPatrol, on behalf of Arbitrum, whose website says it 'makes use of advanced LLM scanning' for 'brand protection for leading Web3 companies,'" he wrote.

"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."

[7]

While Sanderson, who did not immediately respond to a request for comment, attributed the error to automation, Nikita Varabei, co-founder and CEO of ChainPatrol.io, claimed the takedown was the result of human error – an employee pasted the wrong URL into a takedown submission form.

"There is a manual portion of the takedown submission," Varabei told The Register in a phone interview. "And there was some human error in one of the people, one staff member that submitted the takedown.

"They had videos set to autoplay. They had a legitimate scam open. They were doing a couple of tasks. And a new video opened, got switched to, while they were trying to get [the scam video] removed.

[8]

"Honestly, it was just human error. There was no automated AI that submitted this takedown in any way. This specific thing wasn't even reviewed by our system. What we see here is the case of human error where we actually need to have more automated checks in place that can prevent this kind of human error."

[9]Fining Big Tech isn't working. Make them give away illegally trained LLMs as public domain

[10]OpenAI: 'Impossible to train today's leading AI models without using copyrighted materials'

[11]Media experts cry foul over AI's free lunch of copyrighted content

[12]Don't open that 'copyright infringement' email attachment – it's an infostealer

Varabei said ChainPatrol is planning to publish a social media post outlining what happened and how the situation can be avoided in the future. "We are in communication with Grant right now to make sure this situation is resolved. We've already retracted the takedown and the video will be coming back shortly."

Indeed, the Bitcoin video is now back online here, we note:

[13]Youtube Video

"With help from YouTube, the video is now restored," said Grant in an update. "I was personally never too worried. I'm fortunate enough to have a large audience and contacts within YouTube."

Varabei, meanwhile, said false positives are rare for ChainPatrol, which looks for brand impersonation. "We find impersonators and we help get them taken down because there's very many, there're like hundreds of thousands of these. We remove [them] and our false positives are in like the single digits essentially," he explained.

The chief exec also said this wasn't a takedown on behalf of a specific brand.

"This video was not taken down on behalf of any particular company," he said. "It was purely that another completely unrelated video that is actually malicious was in the process of being reviewed. And then essentially the video just happened to go to like whatever YouTube selects to be the next video. And [the reporter] copied the wrong link."

False positives, said Varabei, are something his AI-powered company takes extremely seriously, noting that Chainpatrol responded to Sanderson as soon as it saw his post. "False positives are something that should never happen," he said.

False positives are something that should never happen

Alas, false positives do happen. Game site itch.io, for example, in December [14]said its website was taken down briefly as a result of "AI powered" brand protection software.

And this has been an issue for years, since before the current AI craze. An [15]article [PDF] titled "Copyright False Positives," from the Notre Dame Law Review in 2013, notes, "automated enforcement technologies frequently send cease-and-desist letters or DMCA notice-and-takedown requests for non-infringing uses. In theory, such actions would seem relatively harmless given the numerous legal means available for alleged infringers to contest these erroneous claims.

"In practice, however, even if successfully challenged by alleged infringers, the litigation costs involved in correcting enforcement errors impose a burden on creative expressions and the rightful exercise of public rights and copyright exceptions."

Varabei said people often don't understand that ChainPatrol handles millions of scam sites, fake domains, and fake YouTube videos.

"We try to keep an extremely low false positive rate," he said. "We deal with them very quickly. In some cases, human error like this can happen. And we already have a plan of action that we're starting to deploy and develop to make sure that this doesn't repeat." ®

Get our [16]Tech Resources



[1] https://www.youtube.com/@3blue1brown

[2] https://x.com/3blue1brown/status/1876291319955398799

[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Z3y0-Reb0I4Tip_FruA64wAAAAs&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[4] https://support.google.com/youtube/answer/2814000?hl=en

[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Z3y0-Reb0I4Tip_FruA64wAAAAs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Z3y0-Reb0I4Tip_FruA64wAAAAs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Z3y0-Reb0I4Tip_FruA64wAAAAs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Z3y0-Reb0I4Tip_FruA64wAAAAs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[9] https://www.theregister.com/2024/12/22/ai_poisoned_tree/

[10] https://www.theregister.com/2024/01/08/midjourney_openai_copyright/

[11] https://www.theregister.com/2024/01/12/senate_committee_ai_journalism/

[12] https://www.theregister.com/2024/11/07/fake_copyright_email_malware/

[13] https://www.youtube.com/watch?v=bBC-nXj3Ng4

[14] https://x.com/itchio/status/1866017758040993829

[15] https://scholarship.law.nd.edu/cgi/viewcontent.cgi?article=1704&context=ndlr

[16] https://whitepapers.theregister.com/



Dystopia is here

ecofeco

AI plunders everyone's content, meanwhile human creator gets penalized for using their own creations.

It's Brazil meets Idiocracy.

There are more and more of shit like this

IGotOut

I wish I could find the video, but it goes into depth how you can easily take down legitimate music streams.

In short, you create AI slop music and set it to be monitised. You then can get bots to scan for similar sound pieces, issue an automated strike down. You then get a share of the "infringing" works revenue.

Then there are other YouTubers complaining they are getting takedowns for no real reason except a bot has created a copyright complaint and forced their video offline (even some of the biggest streamers have had this).

Then finally you have the scumbags like Teenage Engineering that issue copyright infringment for criticising their overpriced tat, just because you played a note from one of the instruments (yes read their T&C's, it's really scummy).

Copyright is a joke on YouTube, and they just don't give a shit.

Re: There are more and more of shit like this

Anonymous Coward

Watch youtube with Brave-Browser, no ads = no money to youtube - just hosting bills

Re: There are more and more of shit like this

Sorry that handle is already taken.

You don't even need that scamware, just use ad and script blockers with your favourite normal browser.

Utter bullshit!

Yorick Hunt

"There is a manual portion of the takedown submission"

You can train an LLM, you can form a company around your vapourware, but you can't write a simple JS extension for the meatbag's browser?

This is right up there with "I did not have sexual relations with that woman."

A Perfect System

An_Old_Dog

Our system would have been perfect if it weren't for those meddling kids imperfect, flawed humans. We should have LLMs handle the entire process!

Boilerplate in Corporate 'Apologies'

An_Old_Dog

What they write: "We take ---------------- very seriously."

What they meant: "There is no gay sex in the Army. None."

"And by 'none', I mean, 'rather quite a lot.'"

"They are both businesses - if you have given them enough money, I'm
sure they'll do whatever the hell you ask:->"
-- David Welton