The European Parliament has voted in favor of the text of what is set to become the world’s first artificial intelligence (AI) regulations: the EU AI Act.
Following approval of the most recent draft proposal in May, European lawmakers today passed the text of the law with a large majority, thus entering the final step of negotiations with the European Council, which will determine the final look of the law.
Presenting the outcome of the plenary vote on the AI act, Roberta Metsola, president of the European Parliament, said the act will ‘no doubt be setting the global standard for years to come.’ She added that this first set of regulation aims at managing risks posed by AI and developing a lawful use of such systems.
‘I think we can all be proud of ourselves that Europe is leading and will continue to lead in AI legislation,’ Metsola stated. ‘Innovation brings us forward and opens up new possibilities and as legislators we need to seize the opportunity. It is about change. It is about understanding that we cannot afford to remain stagnant, and about not being afraid of the future.’
She stressed the importance of continuing to set ‘clear boundaries and limits’ to AI and pointed out that every advance in technology must go hand in head with fundamental rights and democratic value.
Check, verify, control
The draft of the rules by EU lawmakers ensures AI systems are overseen by humans, safe, transparent, traceable, environmentally friendly and not discriminatory. It also includes a tiered and risk-based approach to regulation as suggested by the European Commission (EC), which prohibits or allows the unrestricted use of AI tools depending on whether or not they are high risk.
Systems such as ChatGPT or similar, widely adopted by the IR community in recent months, fall into the ‘AI with specific transparency obligations’ category and are therefore not considered to be high risk. What this means is that ChatGPT or other generative-AI tools that operate in the same way would need to clearly disclose that their content was generated by AI and the model would have to be designed to prevent it from generating illegal content or publishing any copyrighted data.
‘GPT [models] can generate content that can tell children how to harm themselves,’ said Dragoş Tudorache, member of the European Parliament and co-rapporteur on the regulation. ‘Is it something you want in our society unchecked, verified and controlled without some clarity as to the roles and responsibilities for that source content? My answer, our answer in this house is no.’
Tudorache added that the European Parliament and the EC believe standards are necessary to protect the interests and value of the union.
‘That’s precisely what the legislation does,’ he said. ‘And it doesn’t do it – this is very important to stress – it doesn’t do it by unnecessarily putting barriers for businesses and for creativity and innovation, because it limits very much the scope of those who will have to comply only for high-risk applications.
‘We have decided not to regulate the technology itself, but to look at the uses of the technology and categorize [the risks it poses].’