EU Draft Generative AI Rule Requires Listing Copyrighted Training Material
Generative AI and synthetic media developers will need to compile a list of every piece of copyrighted material used to train their models should the European Union cement a newly advanced draft of the AI Act. EU lawmakers have spent two years working on the AI Act but added the generative AI rule just ahead of the new vote, which sent the Act to the trilogue stage of debate among EU legislators and member states.
Generative AI Regulation
Members of the European Parliament (MEPs) set the new rules under the category of General Purpose AI, which lays out the requirement of disclosing any copyrighted materials used to train AI models regardless of their output form in text, image, or video. OpoenAI, Stability AI, and others would need to make a list of every bit of training that involved material protected under copyright. The new rule only came up for discussion in the last couple of weeks and could have led to an outright ban on any generative AI model using copyrighted material. The transparency rule is a compromise choice.
The sudden inclusion of generative AI rules for the AI Act after two years comes amid a flurry of European regulatory investigation and action into the technology. The Italian Data Protection Authority ordered OpenAI to block ChatGPT in the country last month, later issuing a set of demands that OpenAI had to meet by April 30. OpenAI uncovering and resolving a ChatGPT vulnerability that could have allowed some users to see the titles of conversations others had with the chatbot sparked the investigations, but Italy had many other requirements to allow ChatGPT back into the country. OpenAI made enough changes to be allowed back into Italy over the weekend. OpenAI now offers a form that users in the European Union can fill out and submit to have personal data removed, a new age verification tool for when people in Italy sign up, and a help center article explaining ChatGPT’s data collection system and steps to take if there is a privacy concern.
Investigators in Germany and other countries have begun putting together reports on ChatGPT as well, though without banning the generative AI chatbot. Italy also requested the European Data Protection Board (EDPB) set up an investigation, which it has done, though that’s unrelated to the AI Act. the EDPB is focused more on supplementing the General Data Protection Regulation (GDPR) rules instituted a few years ago.