October 30, 2023

Spain fires the starting gun for its AI sandbox

Spain rules out financial help for AI startups taking part, despite calls from tech lobby group

Spain has ruled out cash for AI startups taking part in a national initiative to help them adapt to the EU’s AI Act.

The Spanish government has begun reaching out to AI tech companies to take part in its new AI sandbox, the first attempt by an EU country at helping founders adapt to the legislation.

The new rules, which some AI startup founders have warned might persuade companies to relocate to the US to avoid the extra bureaucracy, will come into force in 2025. 


Spain’s AI minister Carme Artigas tells Sifted that 52 organisations, including startups, had “expressed interest” in taking part in the first iteration of the sandbox, which will last six months. 

She says 137 independent experts have also volunteered to take part in the project, which she describes as “a great success”. 

But some startup founders who have spoken to Sifted have warned that the burden of dealing with the paperwork could be time consuming and costly — and could even require them to hire more staff to cope with the extra work.

“When you get involved in something like this you need to allocate resources, hours, effort, and it’s true that if we had some financial support we would be able to hire one more person to help us deal with the paperwork and the communications with the government,” says Nico de Luis, chief executive of Madrid-based Shakers, an AI-powered recruitment startup which applied to take part in the exercise when the application process kicked off two weeks ago. 

Tackling risks

The sandbox is targeting AI companies deemed to be working on foundational models and high-risk AI systems — those that negatively affect safety or fundamental rights as defined in the EU draft legislation — as well as other companies headquartered or with an office in Spain that are using AI.

The AI Act draft, which is still being negotiated in Brussels, would place a number of obligations on developers of high-risk AI systems, including requiring companies to carry out a fundamental rights impact assessment and ensuring that humans operating them are made aware of the risk of confirmation bias.

The regulation would consider the rest of AI systems as either “low-risk” or “unacceptable risk” — with the latter being applied to use cases that will be outright banned, such as those using manipulative or deceptive techniques to distort someone’s behaviour, and facial recognition technology in public places.

To help smooth companies’ adoption of the new EU rules, the Spanish Ministry of Economic Affairs and Digital Transformation will provide participating businesses with legal support on how to comply with the regulation from a new AI advisory council, and a chance to send back their views. 

The aim is to produce implementation guidelines that could be applicable in other EU countries. 

A pan-European AI sandbox

The European Commission has said other EU member states might join the Spanish pilot in what could potentially become a pan-European AI regulatory sandbox.

Santi Molins, co-CEO and cofounder of Barcelona-based TalentFY, which uses AI algorithms to score candidates for a job, also thinks startups like his would feel the strain. “We don’t have any lawyers in the company. We’re just eight people,” he says.


Antonio Tripiana, tech lead at Barcelona-based fintech Deale, which has applied to take part in the sandbox, agrees. Deale’s 30-strong team uses AI to match investors with companies, and bring together all their relevant data.

“It’s difficult enough to develop AI models, integrate them with our platform and then on top of that think about all the regulatory aspects,” Tripiana says.

Justo Hidalgo, chief AI officer at Spain’s tech lobby group Adigital, predicts this will be a tough, months-long and expensive process, especially for startups, and suggests a government subsidy to prevent companies facing such high costs — which he says could push startups to take less risk or move outside of Europe.

“They are all analysing what the sandbox might mean, how it might affect them and how it could help their business,” he says. “Companies must perceive a value in taking part [...] [The extra work] can create a problem with investors, because founders would have to ask for money not just for their technical and commercial teams, but also for a regulatory team.”

He also suggests something akin to a badge of quality if they show they’re fully compliant.

But Artigas has ruled out government financial support for startups taking part, arguing that granting subsidies would be “discriminatory” and that participation in the sandbox will already give companies a competitive edge.

“The sandbox is voluntary and it will give a competitive advantage to those companies, which will be able to get ahead of the market before the rules come into force,” she says. 

“What is required from them is to allocate people and time, if someone can’t allocate them, they shouldn’t take part in the sandbox.”

Cristina Gallardo

Cristina Gallardo is a senior reporter at Sifted based in Madrid and Barcelona. She covers Europe's tech sovereignty, deeptech and Iberia. Follow her on X and LinkedIn