Effy Vayena, an expert in technology law, explains in an interview what the new AI rules will bring to consumers – and what points worry her.
Effy Vayena works as a professor at the Institute for Translational Medicine at ETH Zurich.
Karin Hofer / NZZ
The EU passed the world‘s first comprehensive law on artificial intelligence on Wednesday. It introduces standards that all AI products in high-risk areas must comply with, including applications in recruiting, assessing creditworthiness, in the judiciary, but also in schools.
We spoke to Effy Vayena about it. She is a professor of bioethics at ETH Zurich and researches technology and governance, especially in the area of health. As a delegate for digital transformation and governance, she also advises the ETH school management in this area.
Ms. Vayena, the USA leads the way in innovation, the EU in regulation. Is this a good move for Europe?
It’s not entirely true that the US has no regulation. A lot is happening there, President Biden has issued an order, member states have passed laws. The truth is that they do not go as far and are as broad as the EU’s AI Act. As with data protection, the EU is trying to lead the way and set the tone for how companies should behave.
So is the AI law good news or bad?
It is good that attempts are being made to establish guard rails. But the devil is in the details. How is the law implemented, what standards are established? And how well will it fit in with the technical innovations that are coming our way? This will have to be watched.
The text of the law is there, why can’t we yet assess what it means for companies and consumers?
It depends on how the new rules are implemented. The EU is setting up a new institution, the AI Office. This will translate the law into industry standards together with authorities and private standardization organizations. A lot of detailed technical work is required. How well this all works remains to be seen. It is clear that companies have to do a lot of documentation work; for example, for certain risk categories they have to present plans for how they handle data.
In the future, AI products such as delivery robots will have to meet strict requirements in order to be able to deliver parcels in the EU.
Leon Neal / Getty
This is a lot of effort for startups and smaller companies. Does this threaten the startup scene?
This is a serious concern. The EU is also aware of this and has promised to be more lenient with startups. But I think it’s hard to prevent some smaller players from letting bureaucracy deter them from setting up a company at all. The medical device industry, for example, is already heavily regulated. The new law brings with it further requirements. For large companies that have enough resources, keeping up with the new requirements is less of a problem. This can further strengthen their market power.
Does all this bureaucracy do anything for consumers?
I hope so. For consumers, stronger protection is a good thing, especially in the context of technologies that are spreading extremely quickly. The key question will be whether the new regulations and the new system of standards and control authorities will operate efficiently and effectively in the broader legal ecosystem. Can they block the “bad” innovation and enable the good one? Or do unwanted applications find loopholes again?
What would you have done differently if you could have decided?
This method of broadly regulating the entire market has now become established in the EU. This also has to do with the history of the EU and the challenge of reconciling many interests under geopolitical pressure. In general, I am in favor of regulation that focuses not so much on a specific type of technology, but on how we humans interact with technology and how our rights are protected. Therefore, the general requirements should be broad and the details should be defined in the specific sectors.
They said it would be something to watch. But will consumers actually feel the rules? It is often difficult to say what a law actually achieved.
When I get on a plane or get an MRI scan, I trust that these things are safe because there are facilities to take care of them. This should also be the case with critical AI products. Regulation must support those institutions that take care of it. This requires good regulations: enforceable, lean, smart. I hope that the EU law will be implemented in this way. Because we want good technology to be developed.