In the European Union entered due to the first stage of regulation of AI: from February 2, systems are prohibited that carry an “unacceptable risk” for society. This is part of a large -scale act on artificial intelligence adopted in March 2023 after long discussions.
EU divided II systems into four categories of risk. The minimum (for example, spam filters) and limited (chat bots) do not plan to regulate. High -risk decisions, including medical systems, will fall under strict control. But II, who are considered “unacceptable” are now banned completely.
What is the ban?
The law prohibits several types of AI:
- Social rating systems that evaluate people according to their behavior.
- Algorithms that manipulate people’s solutions with hidden methods.
- Technologies operating human vulnerability (for example, age, disability, financial position).
- predictive systems that evaluate the probability of committing crimes in appearance.
- Biometric and determining personal characteristics, such as Sexual orientation.
- systems of real time that monitor biometrics in public places for the police.
- Analysis of the emotions of employees and students.
- Expansion of person recognition based on the Internet scanning and surveillance cameras.
companies that violate the law can receive a fine of up to € 35 million or 7% of the annual turnover. At the same time, the prohibitions relate to all firms working in the EU, regardless of their headquarters.
Although the period of entry into force of restrictions is February 2, fines will begin to operate only in August. By this time, the regulatory authorities and the mechanism of their work should decide.
In 2023, more than 100 companies, including Amazon, Google and Openai, voluntarily signed AI Pact, undertaking to observe the principles of the act in advance. However, giants such as Meta, Apple and Mistral AI ignored the initiative.
Are there any exceptions?
Yes. The law allows the use of prohibited technologies in exceptional cases:
- The police will be able to use biometric systems to search for missing people or prevent a threat to life.
- Determining emotions is permissible if it is used in medical or medical or Therapeutic purposes.
Nevertheless, the European Commission promises to submit accurate rules and clarifications no earlier than 2025.