Companies deploying generative artificial intelligence tools such as ChatGPT will have to disclose any copyrighted material used to develop their systems, under an early EU agreement that could set the world alight. Paving the way for the first comprehensive law governing technology.
According to Reuters, the European Commission began drafting an “artificial intelligence bill” two years ago to regulate the emerging technology, which has ushered in a boom after OpenAI released ChatGPT. Members of the European Parliament have agreed to move the draft forward to the next stage, a tripartite dialogue, when EU lawmakers and member states hammer out the final details of the bill.
Under the proposal, AI tools would be categorized according to their perceived level of risk: from minimal to limited, high and unacceptable. Possible areas of concern include biometric surveillance, the spread of false information or discriminatory language. While high-risk tools will not be banned, those who use them need to be highly transparent in their operations.
Companies using generative AI tools must also disclose any copyrighted material used to develop their systems. The clause is a late addition drawn up over the past two weeks, a person familiar with the matter said. Some committee members initially proposed an outright ban on using copyrighted material to train generative AI models, but that proposal was dropped in favor of transparency requirements, the person added.
Svenja Hahn, deputy member of the European Parliament, said: “With conservatives wanting more surveillance and the left fantasizing about over-regulation, Parliament has found a solid compromise that moderately regulates AI, protects citizens’ rights, and fosters innovation. and promote economic development.”