Article 50 of the EU AI Act applies from today, so chatbots must identify themselves and deepfakes must be disclosed. The machine-readable marking rule for generative systems already on the market slipped to 2 December, which is the part most builders actually needed.
Read the full story: EU AI labelling rules go live today, minus the big one →
Transcript
As of today, Article fifty of the E U A I Act applies. If you serve anyone in Europe, a chatbot has to make clear it is a machine. A deepfake has to be disclosed. And A I generated text published to inform the public has to say so, unless a human genuinely reviewed it and someone took editorial responsibility. Get it wrong and you are looking at fifteen million euro or three percent of worldwide turnover, whichever is higher. But the rule engineering teams actually built for is not the one that landed. Machine readable marking of synthetic output does not bind generative systems that were already on the market. Under the deal Council and Parliament struck in May, that slides to the second of December. Read that carefully though, because it is narrower than the relief people are hoping for. A model you launch tomorrow is bound immediately. Only the ones that shipped before today get the extra four months.