AI ethics: Who should be responsible?

30 Haz 14:55

With AI taking on more autonomy, the issue of responsibility is a bottomless pit where every stakeholder points fingers in their own direction. Developers often say, "I only wrote the algorithm; the final decision was made by the model," which is like blaming the compiler: legally valid in some cases, but ethically questionable. Then there are users, who sometimes don’t even realize they’re interacting with an AI, let alone one capable of making decisions. And in the middle, there’s the AI itself—which isn’t a legal entity… yet. The solution isn’t simple because it blends engineering, law, and even philosophy. First, you need real algorithmic transparency—not just publishing a polished paper explaining the model in broad terms. Regulators (the EU is already ahead with the AI Act) are demanding independent audits of critical systems, with full traceability for every decision. This forces developers to realize that every line of code could end up in court. Then there’s the matter of tech insurance. Companies are already starting to require specific policies for their AI systems, just like they do for autonomous vehicles. This shifts part of the financial responsibility to insurers, who will in turn audit the models before covering them. It’s not perfect, but it does mean someone else has to review the code before it goes live. And let’s not forget the end user: education is key. Tools like these shouldn’t be released to the public without clear, mandatory documentation on their limitations. If someone uses an AI system to diagnose cancer in a hospital and it goes wrong, the doctor needs to be able to prove they understood it was just "assistance," not an infallible oracle. In the end, responsibility is shared, but it always starts with those who have the power to change the system: the developers.