Possibly — and the part that already applies is the part people overlook. The AI Act's transparency obligations under Article 50 have applied since 2 August 2026. The widely reported "delay" moved the high-risk obligations to 2 December 2027 and 2 August 2028; it did not move Article 50.
Why this matters to you specifically. Article 50 applies not only to providers of AI systems but also to deployers — the people, companies and organizations that merely use them. If you built something with GLACIER that talks to your users, you are a deployer.
What Article 50 covers, in plain terms: telling people when they are interacting with an AI rather than a human, marking AI-generated content, and disclosure around emotion recognition, biometric categorisation and deepfakes.
The number that focuses the mind: non-compliance with these transparency duties can attract fines of up to €15 million or 3% of worldwide annual turnover.
One more date worth knowing: from 2 December 2026, providers of systems already on the market before August 2026 face a transitional deadline for machine-readable marking of synthetic content.
What this is not. This article is orientation, not legal advice. Whether and how the AI Act applies depends on what you built, who uses it and where. Speak to your own advisor.