Important update. The AI Act may be changing, but businesses should not treat this as a pause button.
The real issue is still operational: can companies prove where AI was used, what was disclosed, who reviewed it and whether the process was under control?
I recently wrote about this from the Article 50 transparency perspective.
Great summary, Luiza. The high-risk obligations defer to 2027/28, but the August 2026 date holds for most of the Article 50 transparency duties. Article 4 literacy already applies, and the new NCII/CSAM ban is December 2026. The deferral is definitely there, but it’s mostly about high-risk AI systems. If you are deploying an AI system which is not high-risk, a lot of obligations are already live or soon to be.
Important update. The AI Act may be changing, but businesses should not treat this as a pause button.
The real issue is still operational: can companies prove where AI was used, what was disclosed, who reviewed it and whether the process was under control?
I recently wrote about this from the Article 50 transparency perspective.
Great summary, Luiza. The high-risk obligations defer to 2027/28, but the August 2026 date holds for most of the Article 50 transparency duties. Article 4 literacy already applies, and the new NCII/CSAM ban is December 2026. The deferral is definitely there, but it’s mostly about high-risk AI systems. If you are deploying an AI system which is not high-risk, a lot of obligations are already live or soon to be.
The biggest win in today’s update is the immediate ban on "nudifier" apps.