Artificial intelligence (AI) is everywhere, but it has largely been unregulated to this point. The European Union is looking to change that with its proposed EU AI Act. The European Parliament took a major step forward in AI regulation at the beginning of June when its members agreed on a draft version of the Act.

The Act was first proposed by the European Commission in 2021, and much has changed with AI since then – especially with the rise of generative AI products. As a result, it’s not unexpected that Parliament’s adopted negotiating position differs from the Commission’s or the European Council’s position adopted in 2022. The differences in the three versions of the draft AI Act may prolong negotiations and the final regulations. However, companies that provide or use AI systems can anticipate the approximate direction of the regulations and should begin preparing.

The AI Act takes a risk-based approach to regulating AI, outlining stricter regulations for AI applications that pose a greater threat to user safety. AI systems that fall into the “unacceptable risk” category of the Act would be prohibited in the EU.

Key changes

The European Parliament’s adopted position goes farther than the Commission’s and Council’s versions, with new obligations for providers and expanded scopes for what qualifies as a high-risk AI system. Some of the key changes include:

• A stricter definition of an AI intelligence system that aligns with the definition from the Organisation for Economic Co-operation and Development (OECD). The draft defines an AI system as a “machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions that influence physical or virtual environments.”
• A broader list of AI systems categorised as “unacceptable risk” that will be strictly banned. That list now includes many remote biometric identification systems, predictive policing systems, emotion recognition systems, and “social scoring” biometric categorisation systems.
• An expanded classification of high-risk systems to include those that could cause harm to people’s health, safety, or fundamental rights, as well as cause harm to the environment, among others.
• Obligations for providers of foundation models to conduct risk assessments for their systems, as well as to register in an EU database.
• New transparency requirements for generative foundation models to comply with, such as disclosing when content was generated by AI, designing AI models to prevent them from generating illegal content, and publishing summaries of the copyrighted data that the models use for training.

The adopted position also addresses generative AI, which was less of a concern in previous versions. Those changes include:

The AI Act now goes to full negotiations for the Commission, Council, and Parliament to determine the final text of the law, which could take through the end of the year. As we’ve seen in other industries, such as with medical devices, the regulations outlined in the AI Act may soon be obsolete given the rapid pace of innovation in AI technology. Lawmakers are taking steps to address this dilemma, but for now the focus appears to be on creating a foundation for regulating AI systems.

Companies in all industries – whether they provide or use AI or not – should be closely following progress of the AI Act. While it likely won’t go into effect until 2026, there are expected to be steep consequences for non-compliance. For example, placing an unacceptable AI system on the EU market would be punished with administrative fines of up to €40 million or 7% of a company’s total worldwide annual turnover (whichever figure is higher) under Parliament’s adopted position.

With AI increasingly viewed as the technology of the future, and with other governments expected to use the EU’s AI Act as a blueprint for their own regulations, companies should take steps now to adjust their operations for AI and its regulations.

To discover how this nascent technology is set to revolutionise consumer and product safety, register for our webinar here. During this exclusive 45-minute session, our panel of experts will share their perspective on AI’s potential impact on business practices across the product lifecycle. Don’t miss out – register here.