EU AI Act Enforcement: We Moved to 35% and the Parliament Vote Explains Why — But the Story Isn't Over
TexTak's forecast that the EU AI Act high-risk enforcement deadline holds at August 2, 2026 sits at 35% — down from where we'd have placed it six months ago. The movement reflects a specific chain of events: the European Parliament's March 26 vote at 569-45 in favor of the Digital Omnibus delay proposal, the Council agreeing its own mandate on March 13, and the Commission itself proposing the December 2027 extension. Today's news confirms the industry readiness crisis that motivates the delay — only 24% of enterprises have begun compliance preparation — while also clarifying that the legislative path to delay still carries real execution risk.
Let's explain the reasoning chain behind 35% in plain terms. A 569-45 Parliament vote is not a close call. When the body that needs to pass the delay is voting 93% in favor, and the Commission proposed the delay, and the Council has agreed its mandate, the political consensus for delay is about as strong as EU legislative processes produce. That's the primary driver of our downward movement. We're not at 15% because the legislative process itself is the remaining risk — the Digital Omnibus still needs to complete trilogue negotiations and formal adoption before August 2. If that process stalls or faces unexpected complications, the original deadline snaps back into force with no transitional mechanism.
The compliance readiness data creates a secondary dynamic worth examining carefully. Only 24% of enterprises using AI in HR have begun formal compliance preparation, with penalties exceeding GDPR at up to 7% of global turnover. This is both a reason the delay is politically attractive and a potential reason the delay is genuinely necessary — a hard August 2 enforcement date against 76% non-compliance would either generate mass enforcement chaos or demonstrate that the regulation lacks teeth. Neither outcome serves the Commission's interests. The readiness crisis is therefore circumstantial evidence supporting the delay forecast, but we want to be clear: low compliance rates don't guarantee a delay happens. They explain why the political will for delay exists, which we already knew from the vote count.
The part of our forecast that deserves scrutiny: we're treating the medical device AI extended transition (August 2027 under existing provisions) as separate from this forecast, which targets the broader high-risk enforcement date. That distinction matters because even if the Digital Omnibus delay doesn't complete in time, some high-risk AI systems in regulated products already have extended timelines. The August 2, 2026 deadline holding would be most consequential for AI in recruitment, performance management, and workforce monitoring — the HR applications the news covers — not necessarily for the medical device sector. Our 35% is specifically about the original deadline holding in force; it doesn't mean 65% of affected companies face a clean enforcement environment if the delay happens, because the legislative uncertainty itself creates compliance ambiguity.
What would move us below 20%: formal trilogue completion and publication of the Digital Omnibus in the EU Official Journal before July. What would push us back above 50%: a meaningful legislative procedural obstacle emerging in May or June that delays the delay — coalition fracturing, a member state blocking maneuver, or a legal challenge to the omnibus approach. We're watching the trilogue calendar, not the political rhetoric. The politics are settled; the procedural execution is what remains uncertain.