(P03) (INCYBER) AI in a quest for trust

Trust is the key to the adoption and appropriation of technology, especially when it comes to artificial intelligence. Deep learning and neural networks make it impossible for the average person to understand the mechanisms used and the results produced by the machine. We therefore need to “objectify” a “trusted AI” with clear, precise criteria. The AI Act, issued by the European Commission and adopted in June 2023, defines it as legal, ethical, robust, and secure. But are these criteria sufficient? Conversely, do they risk hampering innovation? Can trust in these technologies extend to decision-making using systems that will be autonomous in the future? What are the limits to this trust, particularly in terms of ideological bias and compatibility with our personal and private data? Surely the advent of AI, on its own, will end up eroding trust? Or can AI help to strengthen trust in digital technology through its many uses in cybersecurity? Finally, on a practical level, how can we measure trust in a way that is accessible to the general public? Will we be able to certify it?