How will the EU’s Artificial Intelligence Act affect your company?
28.09.2023
With the Artificial Intelligence (AI) Act, the EU is proposing binding legislation regulating the use and development of artificial intelligence. Why is such a proposal necessary? It is even possible to regulate AI? And what is Switzerland doing in this regard?
In April 2021, the European Commission presented the proposal for the AI Act – the world’s first legal framework for artificial intelligence (AI). According to this legislation, AI systems are analysed and classified according to the risk they pose to users. The higher an AI system’s risk level, the more stringently it should be regulated. The AI Act is currently being discussed with the individual EU countries. A final version of the Act should be available by the end of 2023, while implementation is not expected before 2025. What exactly does the AI Act say? What does it mean for Switzerland? And what are the implications for your company?
When is an algorithm considered to be intelligent?
Everyone has been talking about AI at least since ChatGPT was released in early 2023. While some people are delighted about the new opportunities that AI opens up, others fear that it could threaten their job security – and could even spell the beginning of the end of democracy. The reality of the situation is probably somewhere in between.
The term “artificial intelligence” has a very wide definition. Beyond a certain level of complexity, every algorithm becomes a black box as far as humans are concerned. If it is no longer possible to trace the path from input to output, the algorithm is classified as ‘intelligent’. Ergon is committed to promoting open AI with transparent documentation of both algorithms and their decision-making processes. “We understand AI as a variety of algorithms that enable people to use machines intelligently,” explains Wilhelm Kleiminger (Head of Data Science at Ergon). “So it isn’t the machine that’s intelligent, but rather the way in which people use it.” For intelligent use to be truly successful, people need to educate themselves and learn to classify the results correctly.
“Ergon understands AI as a variety of algorithms that enable people to use machines intelligently. So it isn’t the machine that’s intelligent, but rather the way in which people use it.”
AI – the opportunities and the risks
Like any technology, people can use AI for both good and bad. On the one hand, AI promotes the sustainable use of resources. One example of this is the smart home, which enables homeowners to save energy thanks to intelligent heating and shading control. AI-controlled early-warning systems can predict environmental risks, and intelligent algorithms can promote the development of medications.
On the other hand, people are afraid of artificial general intelligence (“AGI”) that is capable of thinking and acting like a human being. While these fears may be exaggerated, the use of AI still poses a genuine threat. If AI makes existing professions obsolete and creates new fields of work, there is a real risk of social upheaval. AI can also be used specifically to manipulate people – for example, through artificially generated photos or video recordings that are not identifiable as such. AI also poses a threat if it uses information about our online behaviour in such a way that we are no longer capable of making autonomous rational decisions. “Deepfakes and manipulation for political or marketing purposes are nothing new in themselves,” comments Mr Kleiminger. “But what makes AI such a threat is its scalability and therefore greater reach.”
What is the EU doing, and how is Switzerland responding?
The purpose of the EU’s AI Act is to regulate the use of AI by sector. This is a sensible approach, since AI systems that are used in critical infrastructure, schools or law enforcement are considered to be high-risk. Accordingly, they must meet stringent requirements. In contrast, systems such as AI-based traditional video games or spam filters pose only a minimal risk from the EU’s perspective. This is where the AI Act does not intervene. But there are grey areas here too. Once AI is used to optimise pay-to-win games, for example, it manipulates human behaviour.
As was also the case with the introduction of the General Data Protection Regulation (GDPR), Switzerland has been rather cautious with regard to regulation. But it is to be expected that a similar law regulating AI will also come into force here. “Swiss companies would therefore do well to address the issue now,” remarks Mr Kleiminger. “Companies that have clients in the EU will have to comply with the AI Act anyway.” The AI Act applies to Swiss providers who place AI systems on the market or put them into operation in the EU – and violations are punishable with severe fines. Accordingly, many Swiss AI providers will already align their products with European legislation during the development process. So the standards set out in the AI Act will be automatically enforced in Switzerland too.
Are you ready?
There will be no avoiding AI in the future. Companies should offer relevant training and hands-on workshops so that employees can learn how to deal with it at an early stage. Such instruction should address both technical knowledge and the ethics of these technologies. This will increase people’s acceptance and willingness to innovate. Companies that are developing their own models create maximum transparency by taking an open approach and thus support acceptance. When considering use cases for AI in-house, companies must face up to the opportunities and risks that such models pose to both people and the environment. This is how to make the intelligent use of these new technologies a success. And, very importantly, the principles of privacy by design and informed consent still apply to the collection and use of data.
If you implement these recommendations now, you have nothing to fear from the AI Act. What’s more, you will help to turn AI into a pure opportunity that benefits us all.
