EU Member States Agree on Groundbreaking AI Regulation

On Friday, member states of the European Union agreed to unprecedented legislation at the global level to regulate artificial intelligence after intense negotiations regarding the balance between innovation and security.

Ambassadors of the twenty-seven countries unanimously announced the political agreement reached in December between the states and the members of the European Parliament, according to the Belgian presidency of the European Union Council.

The European Commission had presented its draft titled “Artificial Intelligence Law” in April 2021.

It gained prominence with the emergence of the “ChatGPT” program developed by the startup “OpenAI” in California at the end of 2022, capable of writing essays, poems, or translations in a matter of seconds, which accelerated the discussions.

Similar to other systems capable of generating voices or images, this technology reveals the immense potential of artificial intelligence to the general public.

However, this technology also presents various risks, including the spread of highly realistic fake images, raising concerns about the significant potential to manipulate public opinion.

While there are regulations targeting artificial intelligence in several countries, including China for example, the European legal framework is distinct in its scope.

European Commissioner responsible for this issue, Thierry Breton, welcomed the “historic and unprecedented legislation at the global level.”

He stated, “The Artificial Intelligence Law has sparked great interest, for valid reasons! Today, the states have agreed to the political agreement issued in December, acknowledging the perfect balance found by negotiators between innovation and security.”

Both Paris and Berlin expressed their commitment until the end to protect startups specialized in artificial intelligence, to ensure that “European champions” in this field will not be prevented from emerging in the future.

Diplomats said that concerns were taken into consideration before finalizing the text, allowing the countries to obtained clarifications regarding its implementation.

German Minister for Digital Technology, Volker Wissing, expressed his satisfaction on Tuesday, saying, “We achieved improvements for small and medium-sized companies, avoided disproportionate requirements, and ensured the ability to compete internationally.”

German Minister of Economy, Robert Habeck, considered this legislation as an opportunity to exploit the huge potential of artificial intelligence, while taking risks into account.

He emphasized the importance of easy innovation, legal clarity for companies, and the need for non-bureaucratic structures in its implementation. However, the technology community appeared more cautious, as the European Affairs Manager of the pressure group “CCIA” in the sector, Bonifaz Duchamp, stated on Friday that “many of these new rules are still unclear and could slow down the development and deployment of innovative applications.”

He warned that the implementation of the legislation “will be crucial” to avoid imposing a burden on competitiveness.

Marianne Turdo Bitaker, of the “France Digital” association, which is active in the digital sector, stated that the legislation “imposes significant obligations, despite some amendments for start-ups and small and medium-sized companies”, expressing her fear of “additional regulatory barriers that will benefit American and Chinese competition”.

In terms of generative artificial intelligence, rules will be imposed on everyone to ensure the quality of the data used in algorithm development and to verify that it does not violate copyright laws. European rules require developers to ensure that the voices, images, and texts produced are clearly identified as artificial intelligence products.

Enhanced restrictions will be applied to “high-risk” systems, especially those related to vital infrastructure, education, human resources, and maintaining the system, as they will be subject to a series of obligations such as providing human control over the machine, technical documentation for the institution, or implementing risk management in the administrative system.

Legislation stipulates certain supervision over artificial intelligence systems that interact with humans, with a requirement to inform users about this. Like current European rules on product safety, the text imposes controls primarily on companies.

The legislation includes a few prohibitions that are related to applications that violate European values, such as citizen classification or mass surveillance systems used in China, or remote biometric identification of individuals in public places.

In regards to this last point, states have obtained exemptions for certain law enforcement tasks, such as counter-terrorism.

The European Parliament still needs to definitively approve the final settlement in the spring, which can no longer be modified.

Specific rules will be implemented six months after adoption, and after two years for other provisions.

(AFP)