As discussed in our previous update on this matter, on April 21, 2021, the European Commission published its proposal for the world’s first comprehensive regulatory framework for artificial intelligence (AI) (the “AI Act”). Since then, the AI Act has moved forward in the legislation process, as the Council of the European Union and the European Parliament formed their positions on the matter. The latest development is the vote in the European Parliament on June 14, 2023, on the suggested amendments to the AI Act.
As we previously mentioned, the AI Act takes a risk-based approach and differentiates between AI systems posing: (i) “unacceptable level of risk” – which are prohibited; (ii) “high-risk” – to which most of the AI Act applies; (iii) “limited risk” – the provisions with respect to which are much less stringent; and (iv) “low or minimal risk” – on which no legal obligations are imposed.
The text proposed by the European Parliament emphasizes the potential impact on democracy, the rule of law and fundamental rights, and proposes to expand the categories of prohibited systems that pose an “unacceptable level of risk” accordingly. For example, the European Parliament suggests banning the placing on the market, putting into service or use of AI systems that deploy purposefully manipulative or deceptive techniques; biometric categorization AI systems that categorize natural persons according to sensitive or protected attributes or characteristics, or based on the inference of those attributes or characteristics; AI systems that create or expand facial recognition databases through the untargeted scraping of facial images from the internet or CCTV footage; and AI systems used to infer emotions of a natural person in the areas of law enforcement, border management, and in workplace and education institutions.
The European Parliament also suggests extending the list of “high risk” systems to include, inter alia, AI systems intended for influencing the outcome of an election or the voting behavior of natural persons, and AI systems intended to be used by large social media platforms for recommending user-generated content.
Additionally, the text proposed by the European Parliament imposes new requirements on providers of generative AI systems (such as ChatGPT or Bard). In addition to the transparency obligation, which was already included in the original draft of the AI Act, developers of generative AI systems will be required, among others, to train, design and develop the foundation model in a manner that ensures adequate safeguards against the generation of content in breach of EU law, in line with the generally-acknowledged state of the art, and without prejudice to fundamental rights.
The AI Act will now move forward to the next legislative steps – the first of which is the completion of the trialogue between the European Parliament, the Council of the EU and the European Commission to reach a compromise text, while the aim is to publish the finalized AI Act by the end of 2023.
Once the AI Act officially enters into force, corporate violators could face significant penalties (which according to the European Parliament’s suggestion may be as high as 40 million or 7% of the company’s annual global revenue, depending on the severity of the violation). It is therefore critical for businesses developing AI or using AI in the EU or in connection with EU customers to understand the implications of the AI Act and prepare for its expected enactment later this year.
Please feel free to contact us with any questions that you have on this matter.
