Last Tuesday, the third draft of the EU’s Code of Practice for AI for general use was published. One of the sections deals with copyright.
In this third draft, the copyright measures have been further diluted compared to previous drafts. The requirements for AI companies are lowered, and the ability to submit complaints is restricted.
Particularly concerning is measure I.2.2, which allows AI services to scrape all pirate sites not recognized by courts or authorities as pirate services. Given the number of pirate services available and the length of court proceedings, AI companies could essentially gain access to all books, video games, music, movies, TV series, and sports broadcasts to train their data on without compensating copyright holders.
The draft contains many vague formulations such as “reasonable efforts” and “appropriate measures” regarding compliance obligations, like respecting rights claims during web scraping to acquire data for model training or reducing the risk of models producing infringing results.
Previous drafts included a requirement for AI services to provide a single point of contact to facilitate communication of complaints from rights holders “directly and swiftly.” Now, it merely states that they should designate a contact point for communication with relevant rights holders and provide easily accessible information about it.
The Rights Alliance believes the copyright measures are too unspecific and vague, which risks undermining the obligations of AI giants who may interpret the vague requirements as having ample room to continue collecting copyright-protected material to train their models.
Publicerad: 2025-03-13