Following an agreement on the regulation of artificial intelligence between negotiators from the European Parliament and the European Council last December, the EU AI Act came into force on 2 August 2024. The European Parliament approved the legislation in March this year.
It should be noted, though, that most of its rules specify a transition period until 2 August 2026 (including the rules relating to general-purpose AI models) but the prohibitions on AI systems that “pose an unacceptable risk” will apply after six months and after 12 months.
“The European approach to technology puts citizens first and ensures that everyone’s rights are safeguarded. With the AI Act, the EU has taken an important step to ensure that the adoption of AI technologies complies with EU rules in Europe.” — European Commission (EC) Executive Vice-President, Margrethe Vestager on artificial intelligence’s many benefits for people, society and the economy.
The EC has also defined the AI Pact voluntarily, whereby it calls on AI developers to voluntarily adopt the main obligations of the AI Act before the legal deadlines expire.
“An AI Office will now be set up to help companies start complying with the rules before they come into force. We have ensured that humans and European values are at the very heart of AI development,” said Internal Market Committee co-rapporteur Brando Benifei (S&D, Italy).
Possible shortcomings
- AI developers have a say in whether their systems are considered high-risk.
- There are exceptions for high-risk systems used by law enforcement.
- There is no ban for AI systems capable of identifying and analysing consumer feelings.
- There is no obligation to audit LLMs by an independent third party.
If you’re unsure whether the law applies to you, or how best to remain compliant, get in touch. We are more than happy to hear from you and more than happy to help.