Sweeping rules to police artificial intelligence in the European Union could come as soon as 2023 — but Spain wants to get a move on.
The country this week in Brussels unveiled a new plan to test the EU’s Artificial Intelligence Act, which seeks to enforce strict rules on technologies like facial recognition and algorithms for hiring and to determine social benefits.
Starting in October, Madrid will set up a sandbox — a closed-off environment where hundreds of companies will be able to test their risky AI systems for law enforcement, health or education purposes, following the rules proposed by the European Commission in 2021 and under the oversight of regulators.
“The development of artificial intelligence is a priority in Spain,” the country’s junior minister for digital Carme Artigas told POLITICO.
Spain has already launched several initiatives in the field of AI. Earlier in June, the labor ministry presented a new tool to enable platform workers to request companies like Uber and Deliveroo to explain what’s behind the algorithms deciding their schedules and rating their productivity. Madrid is also set to establish a new artificial intelligence authority by 2023.
The project seeks to give a headstart to European startups and medium-sized companies, which make up a large part of Europe’s economic fabric, at a time when innovation in artificial intelligence is largely driven by Big Tech firms including Google, Microsoft, IBM and Meta (Facebook’s parent company). Smaller companies have warned that the future European AI requirements could prove really challenging to meet.
In a global race to master artificial intelligence, the EU has been trying to push for the development of responsible AI systems. The goal is to give “confidence to citizens and companies that European AI is safe, trustworthy and respects our values,” Internal Market Commissioner Thierry Breton said on June 27 at the launch of the Spanish project.
Under its new scheme, Spain hopes to convince companies working on AI systems like self-driving cars, hiring and work-management algorithms, and health applications to come under the microscope of regulators so that they can help them to follow the flurry of future rules on the quality of data sets and of human oversight. Regulators would also warn Spanish and Commission officials about potentially dangerous loopholes as well as guidelines for industries and best practices.
Authorities would also train their staff to supervise and understand complex algorithms.
Artigas said the EU’s privacy rules, the General Data Protection Regulation, had caught Spain off-guard by having to translate complex legal requirements in a short time. She said the country was “really concerned” about making sure the upcoming AI rules didn’t similarly throw off regulators or put Spanish companies at a disadvantage
The project could prove tricky, though, since European lawmakers and EU countries in the Council are still negotiating on their versions of the AI law, where many controversial issues have popped up. These include calls to fully ban all facial recognition and algorithms to predict crimes or prison sentences. Lawmakers are also still undecided on the enforcement of the rules and have different opinions on regulatory sandboxes.
But Artigas said the Spanish pilot will include AI companies working on high-risk projects that are not seen as controversial, such as autonomous cars or medical AI, and remain flexible. The project will receive € 4.3 million from the EU’s recovery fund.
In a strategic move, the Spanish government wants to reveal the findings of its AI test in the second half of 2023, when Madrid takes up the head of the Council of the EU and seeks to clinch a final deal on the AI rulebook.
This article is part of POLITICO Pro
The one-stop-shop solution for policy professionals fusing the depth of POLITICO journalism with the power of technology
Exclusive, breaking scoops and insights
Customized policy intelligence platform
A high-level public affairs network