The AI Act and the Risk-Based Approach to AI Use

The AI Act and the Risk-Based Approach to AI Use

The use of artificial intelligence will be assessed according to its level of risk.

High-risk AI systems are evaluated based on their potential impact on health, safety, fundamental rights, the environment, democracy, and the rule of law.

Examples of high-risk AI applications include education and vocational training, essential public and private services (such as healthcare and banking), the administration of justice and border control, as well as democratic processes, where AI must not be used to manipulate elections.

It is now essential to assess and mitigate risks, maintain records of AI use, ensure transparency and, last but certainly not least, guarantee human oversight at all times.

This marks the legal recognition of the principle that natural intelligence must always supervise artificial intelligence.

Citizens will also have the right to lodge complaints regarding AI systems and to obtain explanations about decisions made by high-risk AI systems whenever those decisions affect their rights.

Transparency becomes the foundation of everything. General-purpose AI (GPAI) systems, such as ChatGPT, will be required to comply with specific obligations, including adherence to European copyright legislation. They must also provide detailed reports on the data used to train and operate their models. This reflects the growing emphasis on ethics and accountability in the development and deployment of AI technologies.

The AI Act Regulation remains subject to final legal-linguistic review before its formal adoption, which is expected before the end of the parliamentary term (May 2024).

It will enter into force 20 days after its publication in the Official Journal of the European Union and will become fully applicable 24 months later. However, certain provisions will apply earlier, including regulatory prohibitions (such as restrictions on the use of biometric identification systems and other prohibited AI practices), which will take effect six months after entry into force; codes of practice after nine months; general-purpose AI rules after twelve months; and obligations relating to high-risk AI systems after thirty-six months.

What do you think? Will the law be able to keep pace with the rapid growth of artificial intelligence?