Fast Facts
- The European Parliament voted Wednesday in approval of its AI Act.
- The Act will enter into force gradually over the next two years.
European lawmakers on Wednesday voted in approval of the European Union's landmark Artificial Intelligence Act, the world's first legislative answer to the rapid, unchecked growth of the AI industry.
The legislation will ban certain applications of AI outright, specifically those that "threaten citizens' rights."
This includes biometric categorization systems, untargeted scraping of facial images, emotion recognition in schools or the workplace, predictive policing and AI that manipulates human behavior or exploits vulnerable people.
There are specific law-enforcement exemptions to these rules baked into the regulation, allowing the use of real-time biometric identification systems only in certain scenarios, including missing persons cases and terrorist attacks.
Related: Exclusive: PhotoRoom CEO on the critical ways it isn't like OpenAI or Midjourney
The Act highlights "high-risk" systems — which include systems to be used in employment, in critical infrastructure such as transportation and in democratic processes, among other things — and calls for certain obligations around such systems.
Such high-risk systems will be assessed both before being made available, and regularly throughout their lifecycle.
The Act does not classify generative AI systems like OpenAI's ChatGPT as "high-risk," but will require such systems — classified as general-purpose AI systems — to comply with transparency requirements, EU copyright law and the labeling of AI-generated deepfake images.
Part of these obligations will require the companies behind such systems to publicly publish summaries of the content they used to train their models.
“The EU has delivered. We have linked the concept of artificial intelligence to the fundamental values that form the basis of our societies," Civil Liberties Committee co-rapporteur Dragos Tudorache (Renew, Romania) said in a statement. "The AI Act is a starting point for a new model of governance built around technology."
A total of 523 lawmakers voted in approval of the Act; 46 voted against and 49 abstained.
Read the full text of the AI Act here.
Related: No, Elon Musk, AI self-awareness is not 'inevitable'
What happens next
Though it has been approved, the AI Act isn't law yet.
The legislation must still be formally endorsed by the Council of the European Union — which represents the governments of the member states — and is expected to be formally adopted in May.
Once it is made law, the Act will enter into force in several stages over the next two years.
Six months after the official publication of the Act, the ban on prohibited systems and applications will take effect. The governance of general-purpose AI systems will enter into force one year after the Act's formal adoption; the entirety of the Act will be enforceable after two years.
Fines for violating the Act will range from 7.5 million euros (or 1.5% of global turnover) to 35 million euros (or 7% of global turnover).
"I commend the EU for its leadership in passing comprehensive, smart AI legislation," Christina Montgomery, IBM's Chief Privacy Officer, said in a statement. "The risk-based approach aligns with IBM's commitment to ethical AI practices and will contribute to building open and trustworthy AI ecosystems."
Contact Ian with tips and AI stories via email, ian.krietzberg@thearenagroup.net, or Signal 732-804-1223.
Related: The ethics of artificial intelligence: A path toward responsible AI
