AI: latest developments in the EU institutions

 

European Parliament adopted two key AI reports

On 20 October, a majority of MEPs voted in favour of two reports on AI:

-a report on a framework of ethical aspects of artificial intelligence, robotics and related technologies

-a report on a civil liability regime for artificial intelligence technologies

This adoption follows a favourable vote from the Committee for Legal Affairs (JURI). For more information, you can consult the previous briefing here.

Please find below the final version of these reports.

Report on a framework of ethical aspects of AI
Report on a civil liability regime for AI

Commissionner Thierry Breton’s speech on AI

The parliamentary Committee on Artificial Intelligence in the Digital Age (AIDA) held its second meeting on Monday 26 October. Commissioner Thierry Breton gave a speech (in French here) in front of the MEPs to underline the potential of AI for industrial applications. He stressed that the European Commission has two objectives:

·       Focusing on AI use cases: he is not in favour of regulating a specific technology but he wants to focus on use cases and training conditions of algorithms to prevent abuses.

·       Organising a single market for data, especially industrial data.

Thierry Breton explained that the European Commission is preparing a legislation for the beginning of next year. The approach will be risk-based with restrictions limited to high-risk uses. When use cases will convey an exceptionally high risk (e.g. recruitment), essential requirements will apply, for instance on the quality of the training datasets, on robustness and accuracy. Conformity with requirements will be checked before the system is sold (ex ante), through a conformity assessment, and combined with checks after the product is placed on the market (ex post).

The use of biometric technologies in public spaces, for instance airports and football stadiums, will be given a particular attention. These systems will always be considered high risk, conformity will be checked ex ante and additional requirements could apply.

For systems with minor risks, such as chatbots, the requirements could be limited to clear information to users.

Rules will apply to AI systems developed by European companies or third countries companies. However, Thierry Breton pointed out that “AI is a technology which is too important for us Europeans to depend on non-European companies and algorithms. We need to organise ourselves for European alternatives to emerge”.

The objective is also to develop an ecosystem of excellence in Europe. Thierry Breton plans to attract every year, for the next decade, over 20 billion euros of investment in the EU.

On the question of Europe’s sovereignty, Thierry Breton underlines that Europe cannot miss the wave of industrial data. The aim is to have European data, including industrial data, stored and processed in Europe in line with European standards. European data cannot be subject to extra-territorial jurisdiction. He referred to the future European Alliance for cloud and European alliance for processors.

 

Next steps:

-11 November: the European Commission will present its Data Governance Act. The Act will, among other things, enable the re-use of sensitive public data, such as public hospital data. This new legislation will also foster the development of alternative platforms where users would be in control of their data and the conditions to share them. These alternative platforms would give rise to individual data spaces, like a “digital twin”.

-Q1 2021: the European Commission will present a Data Act to increase fairness in the data economy by clarifying use rights in B2B and B2G contexts.

 

If you have any questions, please do not hesitate to contact Camille Dornier - Policy Manager: camille.dornier@eurosmart.com

Eurosmart
Rue de la Science 14B - 1040 Brussels BELGIUM
Privacy Policy - EU transparency register #21856815315-64
Twitter LinkedIn
Modify your subscription    |    View online