EU Parliament has approved a groundbreaking AI law, dividing artificial intelligence into four risk categories, with some AI completely banned due to perceived threats to human rights. The law marks the first legislation within the EU regarding the handling of AI and is also the first of its kind globally.
With a large majority, the EU Parliament has given its approval to the first-ever legislation within the Union on the management of artificial intelligence.
The reason for the law is the "demands on AI considering its potential risks and impact," according to the European Commission. The new rules classify AI into four different risk categories for how it can be used. Systems such as "social scoring" are considered to fall under "unacceptable risk" and will be banned, as they are seen to threaten human rights.
Parts of the new law are expected to come into force in May and fully in 2026. Companies that violate the law risk fines of up to seven percent of their global turnover. All EU countries are also required to establish supervisory bodies. Citizens should be able to submit complaints if they believe they have been subjected to violations of the new rules.
Here's what the different classes mean:
"Unacceptable risk" refers to when AI is considered to pose a threat to people's safety, livelihoods, and rights and will be prohibited. For example, AI systems for "social scoring."
"High risk" includes AI systems used in critical infrastructure, education, employment, or in connection with migration, asylum, and border control. These AI systems will be subject to "strict obligations" before being released on the market. Law enforcement is considered to fall into this risk level.
"Limited risk" includes, for example, chatbots, with a requirement that users must be aware that they are interacting with a machine.
"Minimal risk" allows for free use, such as with AI-supported video games or spam filters.
