I don't think I have an answer to that. I would say that it's somehow when we're using AI to automate human behaviour or decision-making, then I would say that the guiding principle should be at least as good as what it replaces on all dimensions and so don't make a robot or an algorithm take positions or be your general practitioner or whatever unless it actually works better in all aspects. And what do I mean in all aspects? So that also includes explainability for instance why should I have an AI doctor who more often gives the right diagnosis and the right medication, if I cannot ask my AI doctor for an explanation and if I cannot have the type of dialogue with my AI doctor then I would have with my general practitioner. So somehow don't underestimate the complication of what we humans do in the dimensions of problem solving. And also driverless cars, they should also be better than human drivers. Of course we want to make experiments with AI we want to learn etc. But if you want a principle for adoption it should be that we’re only succeeding when these driverless cars are better than humans and on all aspects of driving. It also means to avoid these situations where driverless cars misinterpret traffic situations that humans would never misinterpret, like these cases with Tesla's pilots that mistook a trailer truck for a road sign. Stuff like that. It could be that Tesla autopilots on average are more precise and make less mistakes than humans but they should not make mistakes in the particular cases where humans would never make a mistake. So it has to be better in all parameters, not just somehow in overall statistics that we look at. Of course this makes it extremely challenging and somehow I'm demanding a lot from AI from saying this, but I think that it's the right thing, because we don't want to AI backlash where people say one day, it's worse than what we had or it's somehow making the world the worst place, that would be so silly. Then it's better to wait and somehow develop things a bit more slowly and more carefully. That would be it. We are humans, it's us who develop these tools. We want to use them together with humans and unless they can somehow live up to that, then we should not use them.
AI governance should not be limited to officials like governments, engineers, or corporations because artificial intelligence has an impact on everyone's daily lives. Ordinary folks can and should influence how AI is built, utilized, and regulated. One important way citizens can participate is through public consultations and surveys, organized by governments or international organizations. These allow people to share concerns about privacy, fairness, or job security, which policymakers can then incorporate into laws and guidelines. Education and awareness also play an important role. Citizens who take the effort to grasp how AI works, even at a basic level, are better prepared to question unfair or harmful applications of technology. Schools, universities, and public workshops can encourage individuals to participate in discussions rather than delegating choices to experts. Informed citizens can also reject inaccurate narratives that characterize AI as misleading or unmanageable. Citizens can impact AI governance through their daily decisions. Choosing privacy conscious services,refusing apps with questionable data practices, and supporting transparent technology businesses all send a significant market signal. When enough customers care about ethics and fairness, businesses are obliged to adjust. In this sense, AI governance becomes a shared duty in which individuals participate actively rather than passively. My suggestion regarding the use of Ai – The focus should be more on challenging the algorithm rather than blindly trusting it.
If I could introduce one global principle to guide AI development, it would be: “Do no harm to humans,” combined with a mandatory human-in-the-loop requirement for high-impact decisions. Discussions about safe and ethical AI often lack clarity because societies do not universally agree on what “safe” or “ethical” means. However, a foundational commitment to preventing harm provides a practical starting point. AI systems must be designed and deployed in ways that minimize physical, psychological, economic, and societal harm. Equally important is maintaining meaningful human oversight. Machines may outperform humans in statistical analysis or pattern recognition, but they operate within programmed parameters. High-stakes decisions, particularly in healthcare, law, public administration, and security should not be fully delegated to autonomous systems. A human must retain responsibility and final authority. Trust in AI systems depends on transparency, accountability, and oversight. People must feel secure that these systems will not act beyond human control. Embedding human supervision into governance frameworks reinforces both ethical responsibility and democratic legitimacy. Technology should augment human decision-making, not replace moral agency.
I would introduce a global guiding principle: Any AI must be demonstrably accountable to humans. Why is that in particular? Because many other problems of AI — from discrimination to manipulation to concentration of power — are exacerbated when systems operate without it being clear who is responsible, who can intervene, and according to what rules decisions are made. A very powerful AI is not only a technical tool, but a social influencing factor. The more it shapes the world of work, education, medicine, administration or public opinion, the more important it becomes that people retain the last legitimate control. Accountability means in concrete terms: AI systems must be transparent enough to be audited; there must be clear responsibilities if damage occurs; and users must have the right to challenge decisions or demand human review. In addition, particularly powerful AI applications should not be developed solely according to efficiency or profit, but should be evaluated according to their benefits for dignity, freedom and social stability. This principle would be so valuable because it is not hostile to innovation. It does not prohibit AI, but sets a framework: performance without responsibility is dangerous, performance with responsibility can be enormously useful. In this way, technology remains a means to human goals and does not itself become a benchmark. In short, if I could only choose one rule, it would be this — No AI without clear human responsibility. It protects trust, democracy and human dignity and creates the basis for AI to enable progress without upsetting society.
If there can only be one rule, I would go for: AI developers have to be liable for the system's output
I wish we could fundamentally keep the military out of it, or surveillance agencies like Palantir. I would also like to forbid the intellectual property limitations on them. AI has been trained on the collective knowledge of humanity, and thus access to its fundamentals should belong to all humans and should be open forever.
Regulate people for bad behaviour not the technology
Complete transparency.
AI development under quality control of training data
