The European Commission has proposed the first European legal framework for the use of artificial intelligence, which would increase the security and protect the rights of people and businesses, while encouraging investment in technological development and innovation. The proposal also includes a ban on the use of artificial intelligence in certain cases and financial penalties for rule violations.
Artificial intelligence has been used for decades; however, technological advancements have led to the development of new capabilities and diversification of its uses. While this has opened up many opportunities in diverse fields such as health, transport, energy, agriculture, tourism or cyber security, it has increased the risks associated with its use as well.
In her the presentation of the legislative proposal Margrethe Vestager, executive vice president of the digital affairs commission, pointed out that it is necessary to create synergies between ecosystems of trust and excellence to ensure efficient and safe use of artificial intelligence. The aim of the proposal is to close the current gap between EU and the US, along with China, in the field of digital technologies.
The commission’s proposal is based on the premise that artificial intelligence is a means, not an end in itself. Therefore, the commission would not regulate artificial intelligence as such, but when and how it can be used. Vestager illustrated this principle by saying: “The higher the risk of use for people’s lives, the stricter the rules”.
The Commission proposes a four-part pyramid structure. The bottom and widest part refers to the use of artificial intelligence without risk or with little risk, such as spam filters.
The next layer involves applications with limited risk such as ticketing or ticketing systems. Regarding the use of artificial intelligence in systems such as chatbots, users would have to be made aware that they are conversing with a machine.
A key component of the pyramid is the third layer. It covers the application of artificial intelligence in areas that could significantly affect our lives such as employment, education and critical infrastructure (e.g. transportation) that might potentially pose a threat to human life and health.
In this case, the commission proposes five strict conditions for its use: the provision of quality data, detailed monitoring of the system, adequate information for users, an adequate level of human control in both design and implementation, and compliance with the highest standards of cyber security and accuracy.
Withing the group of high-risk systems the Commission pays special attention to remote biometric identification. Remote biometric identification relies on automatic recognition of human features such as the faces, keystrokes of voices in public spaces and is in principle prohibited.
However, the legislation defines a number of exceptions which include the search for a missing child, prevention of a concrete and imminent terrorist threat, and the detection, search, identification or prosecution of perpetrators or suspects of serious crimes. In such cases the perpetrator is facing more than three years in prison if found guilty.
These rypes of uses require the approval of a judicial or other independent body. In addition, it must be limited in terms of time, geographical scope and databases under investigation.
For the narrowest category with the highest risk, the commission proposes a ban on the use of artificial intelligence. It considers it an unacceptable risk that threatens the safety and rights of people.
It includes systems that manipulate people’s behaviour, such as a voice-controlled toys encouraging children to behave in a dangerous manner, and systems allowing governments to create social scores affecting peoples lives, for instance, a bad social score of a traffic offender would be taken into accont when he is applying for a bank loan.
In extreme cases, it predicts financial penalties for violators – up to 30 million euros or, in the case of a company, up to six percent of its total annual turnover.
The legislation in the form of a regulation would be adopted in accordance with the ordinary procedures. It will be heard and considered by the EU Council and the European Parliament. After it is adopted, the regulation will be directly applicable in all EU member states.