AI hits the wrong note – artificial intelligence still some way to go when detecting copyright infringement

James Rhodes is a classical pianist. He recorded himself playing Bach in his living room and posted the recording on YouTube as he has done on countless other occasions. YouTube immediately removed the video claiming infringement in 47 seconds of the video. Almost immediately stories were circulating on the internet headed ‘Sony Owns Bach’. As it turns out Sony does hold copyright in a number of Bach recordings, a collection featuring Glenn Gould performances. So, the YouTube claim more accurately put was not that ‘Sony Owns Bach’ but that Sony owns a number of copyrights in recording Bach music and that portions of what James Rhodes played sounded very similar to particular parts of the performance by Bach, which are in the ownership of Sony. Far less sensational but definitely equally concerning to a copyright lawyer. The fact that James Rhodes was actually playing should have been enough to halt any sane person from filing the complaint. But that is the point of the story – no actual person was involved. It transpires the whole process was computerised with the help of algorithms within YouTube’s content ID system.  A crawling bot matched the James Rhodes performance with the works in which it held copyright and wham bam it found 47 seconds of what it identified as copying and automatically deleted the section from the recording. To complete its work the system then sent Mr Rhodes a dispute claim notice and from then on in we all know the story.

Copyright law is currently under review and proposals on the table for vote at the European Parliament have reportedly triggered the fiercest lobbying the EU has ever seen in the world of intellectual property. There are two provisions of the current proposals that are making all the waves one of which is pertinent here. It is a provision that would force platforms such as YouTube to compensate owners for use of content on its site and also police the use of content by making the platform liable for unlawful use of copyrighted material.

There is a vote on these provisions in the European Parliament this week and it could go either way. Meanwhile the further development of the technology to track infringement looks urgent. Watch this space for updates.