EU lawmakers adopt recommendations on Artificial Intelligence

Artificial Intelligence in a Digital Age (AIDA), a special committee set up in September 2020 to analyse the horizontal impact of Artificial Intelligence on society, has concluded with its own-initiative report, adopted on Tuesday (22 March).

The AIDA report had a rocky start, as progressive political groups criticised conservative rapporteur Axel Voss for the report’s overall narrative. It was seen as too focused on international competition, where the EU was inevitably falling behind.

After significant redrafting, the report was adopted with a vast majority in the parliamentary committee while maintaining the original emphasis on the potential benefits of the emerging technology.

International setting

“The EU now has the unique chance to promote a human-centric and trustworthy approach to AI based on fundamental rights that manages risks while taking full advantage of the benefits AI can bring for the whole of society – including in healthcare, sustainability, the labour market, competitiveness and security,” Voss said.

At the international level, the report underlines the importance of the EU to play a role as a standard-setter on the global stage, stressing the need to foster the competitiveness of the European tech scene to ensure the EU can shape the international standards with its values.

At the same time, an initial warning that Europe could become a ‘digital colony’ of China was removed. The focus changed to working with like-minded democratic partners.

“Our future global competitiveness in the digital field depends on the rules we set in place today. And these rules need to be in accordance with our values: democracy, rule of law, fundamental rights, and respect for the rules-based international order,” said AIDA chair Dragoș Tudorache, who is also the co-rapporteur for the AI Act for the civil liberties committee (LIBE).

Potential benefits

A crucial part of the report is the Roadmap for AI, a set of policy recommendations for 2030 on the regulatory framework, single market, sustainability, talent, research, e-governance, health, industrial strategy, law enforcement, cybersecurity, and military use of AI.

The question of ensuring adequate access to data appears several times throughout the text. Still, Voss’ calls to reform the EU’s General Data Protection Regulation (GDPR) were removed upon the insistence of left-to-centre MEPs. The wording was aligned with the European data strategy.

“We managed to get strong language on the access for researchers, start-ups and SMEs to the AI economy, including to the data that trains it,” said Green MEP Damian Boeselager. “Market concentration present today in the data economy must be avoided to extend into the AI economy.”

Lawmakers from both sides of the aisle insisted on boosting Europe’s competitive advantage in research, calling for increased investments in research and development to attract and retain AI talents.

The report also supports the establishment of regulatory sandboxes, allowing the development of innovative AI systems and business models under a regulator’s oversight.

Potential risks

The potential risks of AI was at the centre of an event hosted by EURACTIV last week, which outlined critical uses in the justice system, algorithmic management of workers and border control. In these situations with clear power imbalances, AI systems might lead to abuses and the reproduction of existing biases.

“We need to shake off the image that technology is somehow neutral. It’s not. AI is replicating all forms of discrimination, racism, and bias,” said during the event Laure Baudrihaye-Gérard, legal director for Europe at the NGO Fair Trials.

While the AIDA report focuses on the potential benefits of AI, systemic risks are also mentioned. For democracies, the threat is seen in the power imbalance in favour of digital platforms that could undermine the autonomy of citizens. In the hands of authoritarian regimes, AI systems could lead to mass surveillance, ranking of citizens and restrictions on their free movement.

“Remote biometric identification in publicly accessible spaces requires setting massive technical infrastructure, and once we start to build it, it will exist. It will be like the wind that we cannot stop,” said at the same event Wojciech Wiewiórowski, the European Data Protection Supervisor.

On the military use of AI, notably regarding the development of autonomous weapons, the MEPs called for an international agreement on responsible usage of AI in intergovernmental organisations such as the United Nations or the OECD.

“Only by providing an EU model for AI with the appropriate safeguards for citizens, minimising the risks for abuse and undue monitoring, will we obtain trust from the public,” said progressive lawmaker Brando Benifei, the co-rapporteur for the AI Act in the internal market committee (IMCO).

Broader implications

Several parliamentary officials stressed that the relation between the AIDA report and the AI Act should not be exaggerated. Own-initiative reports only go as far as sending a political message, whereas much tougher negotiations are expected for the AI regulation.

Voss, in particular, was accused by his colleague of trying to link the two files. Still, he managed to get through his text on the risk-based approach, regulatory sandboxes, and AI applications for law enforcement. However, the thorny issue of facial recognition was kept out.

[Edited by Nathalie Weatherald]