EU stalls AI rules
- EU governments and Parliament negotiators ended 12 hours of Brussels talks on 28 April without an AI Act deal, pushing the omnibus rewrite into May. - The fight is over carve‑outs for AI inside regulated products — like medical devices, toys, and cars — before rules bite on 2 August. - That leaves companies planning for the original deadline, because the hoped‑for delay to 2027 or 2028 is no longer assured.
Europe’s AI rulebook just hit an awkward pause. Negotiators from the European Parliament and the Council spent 12 hours in Brussels on 28 April trying to soften parts of the EU AI Act, then walked away without a deal. That matters because the law’s next big compliance date is 2 August 2026, and some of the changes on the table were supposed to push key deadlines back by more than a year. Instead, companies are stuck planning for two futures at once. (iapp.org) ### What were they trying to change? The fight is about the EU’s “Digital Omnibus on AI,” a package the Commission proposed in November 2025 to simplify parts of the AI Act and reduce overlap with other rules. Parliament had already backed its negotiating position in March by a 569–45 vote, including later deadlines for some high-risk sy(iapp.org)er to apply in practice. (prod.iapp.org) ### Why did the talks break down? The hard part is overlap. Everyone more or less agreed on delaying some deadlines, but the institutions split over AI that sits inside products already covered by sector-specific safety laws. Think medical devices, industrial machinery, toys, and connected cars. Parliame(prod.iapp.org) (iapp.org) ### Why does that overlap matter so much? Because this is where “simplification” turns into a real legal choice. One camp says companies should not face double regulation when a product already lives under a detailed safety regime. The other camp says carving too much out of the AI Act would splinter one horizontal law into a bunch of se(iapp.org) with some exceptions, or many sector-by-sector versions hiding inside it. (iapp.org) ### What deadline is now causing the panic? The immediate one is 2 August 2026. Without a final deal in time, standalone high-risk systems in Annex III may still have to meet the AI Act’s original compliance date. The hoped-for rewrite would have moved those systems to 2 December 2027, and AI embedded in regulated products to 2 August 2028. But those dates are still only proposed, not locked in. (iapp.org) ### So should companies assume the delay is dead? Not exactly. Talks are expected to resume in May, and people close to the process still see paths to a compromise. But the safe assumption now is that the original August date remains live until lawmakers actually change the law. That is why privacy and AI governance teams are shifting from “wait for the extension” to “prepare as if no extension is coming.” (iapp.org) ### Who is pushing which way? Some industry voices, especially around industrial AI, want relief from what they see as duplicate compliance. Germany’s Friedrich Merz backed easing restrictions for industrial AI, and Siemens argued the reforms should stop “double regulation.” On the other side, Parliament negotiator Brando Benifei warned (iapp.org)rentak went further, calling the failed talks “regulatory chaos.” (iapp.org) ### Why is this politically sensitive? Because the AI Act was supposed to show that Europe could regulate frontier tech early and coherently. Now the bloc is trying to rewrite parts of that law before some major provisions even fully bite. Critics inside Parliament have warned that the simplification push looks too friendly to la(iapp.org)messy. (iapp.org) ### What’s the bottom line? The news is not that Europe killed its AI law. The news is that Europe failed, for now, to agree on how much to soften it before the next enforcement cliff. Until that changes, boards and compliance teams have to treat 2 August 2026 as real — even if Brussels is still trying to move it. (iapp.org)
