EU Launches First-of-Its-Kind AI Act

The emergence of artificial intelligence has deeply impacted all industries and even life. We see it everywhere. And when it comes to smart cities, we see more and more companies relying on it, mostly vision AI and machine learning, but also generative AI, the potential of which is being considered by companies like Anglian Water. With so many turning to the technology to enable their solutions, HaydenAI, Videtics, Vivacity Labs, and cities using it like Paris, where authorities used computer vision during the Olympics for safety reasons, the EU moved towards regulation. As of August 1st, the bloc has put into force the AI Act with regulatory and legal frameworks for AI. So we spoke with European Commission Spokesperson for Digital Economy, Research and Innovation, Thomas Regnier, to discuss AI regulations within the European Union. Thomas, I'd like to start the interview with the AI Act. Around the world, we're seeing a lack of regulation. Like for example, in the US, there's none at all. Why did the EU decide to regulate and how? Well, this is a story that dates back to 2018, when we come up with the white paper on AI. So basically, how we want AI models and systems to develop in the EU. So basically, the AI Act is the first legislation on AI worldwide. So it's quite unprecedented. And our approach to AI is basically to, of course, embrace all the benefits that AI has, both for users and for businesses in the EU, but also to try to tackle the risks and, especially the societal risks that come with this new technology. You approach this from the perspective of risk, and we know that 70% of AI in the EU poses no threat, like spam filters on your email. So they won't be checking that. The market surveillance authority will be checking. But then there's transparency risk, for example chatbots, they need to be properly labeled so that people know they're not talking to a person. Then there's high risk and unacceptable risk, which would be social scoring or real time face recognition. It seems like there are many risks and they're coming from many sides. So how will you control that companies are complying with the AI Act. Basically, what we're asking to be sure that these providers respect our provisions in the AI Act before placing a high risk AI system on the EU market, providers are subject to a conformity assessment. So this will allow them so the providers to demonstrate that their system basically complies with the requirements for trustworthy AI, including data quality. How you train these models, etc. They will also have to comply with, this, there are safety components of products, this is also another type of legislation they will need to comply with this one as well. And then they will have to do quality and risk management. and assessments, they will be subject to audits etc.. So basically all of this they will have to register in the database. So they have a certain as you can see amount of criteria they will have to comply with before being able to be placed on the market. So the EU is the regulator. But you can also help innovation. You don't create the tools but you can you can help the people who actually create them. So there's now regulation in place and the EU invests over €1 billion per year in AI. But how else is the EU helping AI companies improve or advance their AI and encourage talent to stay in the EU and not go abroad? So we are, as I said, within this legislative framework set out by the AI Act, also investing quite a lot. And then we have, as I said, also before, launched our AI innovation package in, January of this year, where we come up with additional funding. and we also opened up our supercomputers to AI startups to be able to train their models. Now, this AI package, also provides, dedicated financial support. to generative AI So we will continue, as I said, to increase our annual spending on AI. And, this new package that we launched in January comes up with an additional €4 billion. So voila, this is budget that comes directly from EU funded programs. My final question for you, Thomas, is, is there an AI smart city project that the EU is investing in to help reach community goals? Maybe what I can give you as an example is something that was launched not so long ago, a month ago. It's the, digital twin Destination Earth project. So the digital twin of our planet, this was also done through AI. What is the goal behind? It is that thanks to this digital twin of the planet, which is now accessible to researchers, to users, etc., we can now predict potential future natural disasters. So you could, for example, say like if now the temperature of the planet is two degrees higher, how would, the level, of the oceans look like? Regulation can both hinder and advance technology. It can hinder it due to excessive cost delaying advancements. Apple, for example, announced that they will not be launching some AI features in Europe due to regulations from the EU. But at the same time, it can help smaller AI startups thrive in a world powered by the Big Five and guidelines allow for a fairer standardization. The unique regulation in the EU will be crucial for smart city solutions. For example, for those with vision AI, as privacy will have to be ensured. However, regulation when it comes to technology, especially one related to the online world will be hard to control with over 5.35 billion internet users around the world and considering only the EU has these regulations. So maybe in the future there will come a time in which, to analyze and control AI, the EU will need to rely on AI.