AI regulation What the proposed Brussels law will mean

The EU's proposed AI law was endorsed by the European parliament on Wednesday, and is a milestone in regulating the technology. The vote is an important step towards introducing the legislation.

- Dan Milmo, Alex Hern

It is now expected to be rubber-stamped by a council of ministers, becoming law within weeks. However, the act will come into force in stages, with a cascade of deadlines for compliance over three years.

"Users will be able to trust that the AI tools they have access to have been carefully vetted and are safe to use," said Guillaume Couneson, a partner at the law firm Linklaters.

The bill matters outside the EU because Brussels is an influential tech regulator, as shown by GDPR's impact on the management of people's data.

How does the bill define AI?

 The law describes the AI technology it regulates as a “machine-based system designed to operate with varying levels of autonomy”, which covers tools such as ChatGPT. This system may show “adaptiveness after deployment” – ie it learns on the job – and infers from the inputs it receives “how to generate outputs such as predictions, content, recommendations or decisions ”. This definition covers chatbots and also AI tools that, for instance, sift through job applications.

The law bans systems that pose an “unacceptable risk”, but it exempts AI tools for military or national security use, issues that alarm many tech safety advocates. It also does not apply to systems for use in scientific research .

How does the bill tackle the risks ?