I really like the per error idea - allows for large cases to have no barrier to contest but for the whole thing to not become like patent trolling. Maybe also some sort of threshold for mass cases, like a progressive taxation:

If you're right 95% of the time, you get a discount. If you're wrong 50% of the time, you pay the full fee.

Who would be in charge of setting these?

Or what about a system like they use in Finland to deal with speeding tickets - the fine is based on your income. The local paper has a weekly list of top offenders, which is amusing to say the least.

Sorry to hijack the thread slightly, but is only to introduce some visibility to the fact that this particular issue is nothing to do with the EU, or content ID, or anything like that.

Google is blackmailing the Blender foundation to try and get them to enable adverts.

https://www.blender.org/media-exposure/youtube-blocks-blende...

edit - and Blender have now got https://video.blender.org/ in testing already. I didn't think they'd be the ideal people to issue ultimatums to on video hosting, given their technical specialities.

A situation like this triggered the development of git, (after the licence to use bitkeeper was withdrawn from the linux kernel team) so I will be watching the fallout with interest.

Apparently Blender had more than 2 years to sign an updated contract - a contract which every partner had to sign to comply with updated legal regulations. Their existing contract would not work anymore (legal reasons?).

The Blender team never did that, and also never reacted to any youtube support mails.

That doesn't map onto the Blender account of what happened.

They said YouTube was requiring them to enable ads, but the ads checkbox was grayed out. Among other issues. Not that they wanted to enable ads on a not for profit site. But just taking this one issue, it's not clear how they could have possibly complied with Google's requirement, when Google keeps the checkbox UI element grayed out and unchangeable.

> Even a tiny fixed fee of $50 per error might be enough

If the fine went to the person who uploaded the work, this would be good because it would force the filter to be robust in the face of adversarial attacks.

That would create a cottage industry of borderline-infringing works designed to trigger the filter but win in court.

I.e. they would trigger bad copyright takedowns that weren't legally justified, forcing companies to stop doing those.

If it's possible to trigger those intentionally, then it's likely that it's also happening accidentally, and this policy would make that stop.

Possibly, but machine generating edge cases for the purpose of collecting fines feels like it should also be wrong.

True, but those edge cases have to convince a court made up of humans, who will probably have similar feelings. That might be a decent check on abuses.

I’m no lawyer, but deliberately performing adversarial attacks to get rich from being automatically and incorrectly fined feels like it ought to be somewhere between fraud and hacking (in the Computer Misuse Act sense rather than thr hackernews sense).