Artificial intelligence regulation

Staying competitive and using artificial intelligence safely: The EU's new Artificial Intelligence Regulation sets out new guidelines for this. All the news on the EU AI Regulation What is the AI Regulation? The EU Artificial Intelligence Regulation establishes a uniform framework for all AI systems. This is intended to regulate the technology transparently, safely, impartially and ethically in future. The EU Commission has been working on the AI law since 2018 and a first draft of the AI regulation was finally presented in April 2021. The European Union plans to use a total of around one billion euros from digital programs for AI, and the EU directive stipulates that private and public players in the digital economy should work together for the purpose of innovation. In future, AI centers of excellence are to be created. Cooperation with research in the fields of AI and robotics is also to be promoted. Why was the AI Regulation adopted? On the one hand, the EU AI Regulation serves to secure Europe's position in international competition. The aim is to strengthen the European single market for artificial intelligence. On the other hand, the EU also wants to create acceptance and trust in AI-based systems. The AI project can be seen as part of a large-scale digital offensive by the European Union, which aims to coordinate crisis responses such as the recent coronavirus pandemic with various EU directives. According to the EU Commission, AI projects that have already been funded have already had a social benefit: For example, useful online tools for checking fake news have already been developed. Artificial intelligence also increases productivity and efficiency for companies. What does the EU AI Regulation contain? The draft AI Regulation provides for a new definition of different types of artificial intelligence. Among other things, it distinguishes between different AI systems based on risk: On the one hand, into AIs with low or minimal risk, AIs with high risk or those with unacceptable risk, which are therefore prohibited. The risk assessment is based on aspects of security, privacy and fundamental rights, and low-risk AIs can be used without specific additional regulations. These include chatbots or spam filters, for example. However, the data generated by artificial intelligence is still subject to the provisions of the GDPR. Which AI systems are considered high-risk? Although high-risk AI systems will continue to be permitted, they will also have to be checked, documented and registered. The EU directive lists the systems that would fall under this AI regulation in future by name. These include AI systems in critical infrastructure, such as in the transport, healthcare or energy sectors, as well as AI systems that perform certain work steps in education, employment, services or law enforcement. All "high-risk" AI systems must therefore commit to transparency in future. At the same time, users should make more informed and independent decisions about the use of such technologies. Which AI will be banned? Artificial intelligence that is considered unacceptably risky includes AI that violates fundamental rights within the European Union. This includes, for example, AI projects that aim to evaluate social behavior by the authorities, so-called "social scoring". Such dystopian AI systems should not become part of society in the future, and toys for children, for example, which could incite them to risky behavior via voice assistance, are also clearly prohibited. The EU AI Regulation thus draws an explicit red line for the use of dangerous AI. How does the AI Regulation regulate remote biometric identification? The search for criminals using biometric surveillance is particularly controversial in this context. The current draft of the AI Regulation prohibits this in principle. The EU Commission cites the protection of personal data and the principle of non-discrimination as reasons for this, although biometric identification via an AI system can be used for user authentication. For example, smartphones can still be unlocked automatically. However, the technology will also continue to be used for border crossings. Who should the EU AI Regulation apply to? The AI law should apply to all players inside and outside the European Union. On the one hand, this means providers of AI systems, such as developers of software applications in the HR sector. On the other hand, it includes users such as companies or the public sector who potentially use high-risk systems and are to be protected by the EU regulation. According to the EU Commission, purely private and non-commercial users are not affected by the AI Regulation. What sanctions are envisaged for breaches of the AI Regulation? The EU Commission's draft AI Regulation provides for severe fines for non-compliance. These include fines of a maximum of EUR 30 billion or six percent of total global turnover for the use of prohibited practices. What is Germany's position on the EU AI Regulation? The German government welcomes the EU's new AI law, in particular its uniform regulations. At the same time, the German Commissioner for the Digital Economy emphasizes the scope for defining the "high" risk group. This is because it also includes biometric identification systems. It must be ensured that innovation is encouraged. The German Ministry of Justice rejected the use of biometric surveillance in public spaces, while German industry was very critical of the EU's AI project. Particularly with regard to high-risk systems, there is a lack of clarity and a risk of double regulation. The regulations are seen as an obstacle to innovation. With an AI strategy, the EU wants to harness the benefits of AI systems for society and the economy. The analyses and overviews of the Europe.Table editorial team provide information on the current status of the draft AI regulation.