EU Council presidency pitches significant changes to AI Act proposal

Slovenia currently holds the rotating presidency of the EU Council. [Alexandros Michailidis/Shutterstock]

The Slovenian Presidency circulated a compromise text on the EU’s draft AI Act, including major changes in the areas of social scoring, biometric recognition systems, and high-risk applications, while also identifying future points for discussion.

The rotating presidency of the EU Council shared a first compromise text on Monday (29 November) to accompany a progress report on the EU’s AI Act.

Scope & definitions

In the progress report, seen by EURACTIV, EU countries restate their exclusive competence in national security matters, and insist that AI systems developed exclusively for military purposes should be taken out of the scope of the regulation.

AI systems developed for the sole purpose of scientific research and development have also been excluded from the scope.

The presidency elaborated the definition of AI systems in order to better distinguish them from classic software programs. AI systems are therefore considered as having the capacity to process data or other sorts of inputs “to infer the way to achieve a given set of human-defined objectives through learning, reasoning or modelling,” the compromise reads.

AI providers are now defined as an individual or organisation “that has an AI system developed and places that system on the market or puts it into service.” Providers will have the responsibility to ensure compliance with the regulation’s requirements.

A new category of ‘general purpose’ AI system has been added, which is not to be considered within the scope of the regulation unless the system is put under a trademark or integrated into another system subject to the regulation.

EU launches AI blueprint in bid to become world leader

The European Commission has presented on Wednesday (21 April) a long-waiting ‘AI package’. The proposal is the first-ever attempt to regulate AI and has been developed as part of a broader ambition to make Europe a global leader in the field by being the first to set clear guidelines.

Social scoring

The Commission’s proposal included a ban on AI applications that are considered to have unacceptable risks. One of these is social scoring, a practice initiated in China that is seen as promoting mass surveillance.

The presidency is now proposing to extend the ban on social scoring from public authorities to private entities. Moreover, the definition of prohibited use has also been extended to include exploiting a “social or economic situation.”

These changes might have far-reaching implications for the financial sector, as for instance interest rates for loans are currently calculated based on the likelihood of repayment.

The use of AI systems for the estimation of insurance premiums was also included as a high-risk system.

Commission yearns for setting the global standard on artificial intelligence

The European Commission believes that its proposed Artificial Intelligence Act should become the global standard if it is to be fully effective. The upcoming AI treaty that is being drafted by the Council of Europe might help the EU achieve just that.

Biometric recognition

The biometric identification systems covered in the legislation are no longer defined as ‘remote’, but any system that leads to the identification of persons ‘without their agreement’.

The possibility of using biometric identification systems in real-time has been extended to actors that are not law enforcement authorities but are collaborating with them. The reason for employing these systems has been extended to protect critical infrastructure.

Biometric systems can only be used prior to the approval of the judicial authority. In case of urgency, the initial proposal provided that the authorisation might also be requested ex-post.

By contrast, based on the new text, the authorisation must “be requested without undue delay during its use, and if such authorisation is rejected, its use shall be stopped with immediate effect.”

Facial recognition technologies already used in 11 EU countries and counting, report says

Law enforcement authorities in 11 European countries are already using biometric recognition systems in their investigations and eight more are to follow, a new study points out, warning of the technology’s impact on fundamental rights.

High-risk systems

The AI Act introduces specific obligations that pose a high risk in terms of health, safety and fundamental rights. In its proposal, the Commission identified eight high-risk areas, which cannot be changed but only further defined in the future.

The most significant change in the list of high-risk systems is the inclusion of digital infrastructure meant to protect the environment, notably “AI systems intended to be used to control emissions and pollution.”

In the area of law enforcement, the subcategory of crime analytics has been removed.

The compromise text provides that the European Commission will have to assess the list of high-risk systems every two years, together with the list of AI techniques and approaches covered by the regulation.

Study warns of compliance costs for regulating Artificial Intelligence

The EU’s forthcoming regulation on Artificial Intelligence could cost the bloc’s economy up to €31 billion over the next 5 years and cause investments to shrink by as much as 20%, according to a study published on Monday (26 July).

Outstanding issues

The progress report anticipates a number of areas that are expected to require further discussion.

The requirements for high-risk systems are singled out as being vague and requiring practical guidance to facilitate business compliance. The examples given relate to how the quality of data and transparency obligations might be fulfilled in practice.

Moreover, several EU countries have stressed that requiring data sets that are complete and free of error might be largely unrealistic. “A number of delegations stressed that while this should be the case to the greatest extent possible, it should not be an absolute requirement,” the report reads.

Several member states also emphasised the complexity of the value chain, “where boundaries between different actors are not always clearly delineated.” As a result, the distribution of responsibilities might need to be re-evaluated to better reflect the reality of AI value chains.

The excessive administrative burden for SMEs was a recurrent theme in the discussions, an issue that was also raised at the last summit of EU heads of states.

Concerns were also raised as to the relation of the AI Act with other EU legislation in order to avoid conflicting legislation, notably in terms of privacy, law enforcement, product safety and other sectorial legislation.

EU leaders push for ‘innovation friendliness’ in digital policy

EU leaders stressed the need to make digital regulation more innovation-friendly, but the digital agenda remained at the margins of the European Council summit.

[Edited by Frédéric Simon]

Subscribe to our newsletters