EU's AI Act 2023 changes (compared to 2021)

The most recent update to the AI Act was on June 14, 2023, when the European Parliament adopted its negotiating position on the Act. The European Council is now expected to adopt its negotiating position in the coming months. Once both the Parliament and the Council have adopted their negotiating positions, the two sides will enter into negotiations to reach a final agreement on the Act.

So, what changed between 2021 and 2023 in the AI Act. It can be summarized as more stringent rules and definitions are introduced. The main changes made to the AI Act between 2021 and 2023 are:

• A new definition of AI: The Parliament's definition of AI is broader than the original proposal, and it includes systems that are not currently considered to be AI, such as facial recognition systems.
• The inclusion of a new risk category: The AI Act now includes a new risk category for "very high risk" AI systems. These systems are subject to the most stringent requirements under the Act.
• New requirements for human oversight: The requirements for human oversight of high-risk AI systems have been strengthened. For example, the Act now requires that human oversight be "continuous" for very high risk systems.
• New requirements for transparency: The requirements for transparency of high-risk AI systems have been strengthened. For example, the Act now requires that developers of high-risk AI systems provide users with information about the system's training data, its algorithms, and its potential biases.
• New requirements for robustness: The requirements for robustness of high-risk AI systems have been strengthened. For example, the Act now requires that developers of high-risk AI systems take steps to prevent the systems from being hacked or manipulated.
• The expansion of the scope of the Act: The scope of the Act has been expanded to include AI systems that are used in the public sector. This means that the Act will apply to AI systems that are used by government agencies, such as the police and the military.
• The introduction of new enforcement mechanisms: The Act introduces new enforcement mechanisms, such as the ability of the European Commission to impose fines on those who violate the Act.

On 12th June, Wednesday, the European Parliament adopted its negotiating position on the Artificial Intelligence (AI) Act with 499 votes in favour, 28 against and 93 abstentions ahead of talks with EU member states on the final shape of the law. The rules would ensure that AI developed and used in Europe is fully in line with EU rights and values including human oversight, safety, privacy, transparency, non-discrimination and social and environmental wellbeing.