Oversight of compliance with the rules on artificial intelligence is carried out by designated authorities at both national and European level, which can require documentation, carry out inspections and impose sanctions for breaches. The system is tiered in a manner similar to the obligations themselves – the use of prohibited practices is assessed most strictly, followed by breaches involving high-risk systems, and finally the failure to meet information obligations. The obligations take effect gradually over transitional periods, so it is worth tracking what already applies and what is still to come. Practical preparation does not consist of studying the sanctions, but of what the authority will actually ask of you: a register of the systems used, their classification by risk, documentation, and evidence of oversight and training. Check the current wording and deadlines against the applicable regulation.
See also: Prohibited AI practices, Register of AI systems, AI risk classification.
