The European Union has revealed steering on what constitutes an AI system underneath its new AI Act. The chance-based framework for regulating purposes of synthetic intelligence got here into drive final summer time — with the primary compliance deadline (on banned use circumstances) kicking in final weekend.
Figuring out whether or not a specific software program system falls inside the act’s scope will probably be a key consideration for AI builders — with the chance of fines of as much as 7% of worldwide annual turnover for breaches. The Fee’s 13-page tips are more likely to be intently parsed by firms.
That mentioned, as with E.U. steering on prohibited makes use of put out earlier this week, the recommendation is non-binding. The Fee additionally provides that the steering is “designed to evolve over time and will probably be up to date as crucial, particularly in gentle of sensible experiences, new questions, and use circumstances that come up”. Given how fast-paced the AI area might be, the duty of understanding how the legislation applies is more likely to stay a piece in progress.
“No computerized willpower or exhaustive lists of programs that both fall inside or outdoors the definition of an AI system are potential,” the E.U. warns in its recommendation doc. Set your expectations accordingly.