A draft proposal seen by the Financial Times shows the EU is planning to hand companies a generous “grace period” and delay fines for violations, giving AI developers time to catch up or wriggle out of compliance. Some rules, which were supposed to kick in for high-risk AI systems by August 2026, may now be postponed by another year, while transparency obligations could be shelved until 2027.

This “simplification package,” up for discussion on 19 November, is being sold as an effort to boost EU competitiveness against the US and China. But what it really shows is Brussels absolutely bricking it over how Donald Trump might react if they dare cross Big Tech’s lobbyists.

A senior EU official confirmed the bloc has been “engaging” with the Trump administration on tweaks to the AI Act and other digital laws. Apparently, there’s now genuine fear that tightening the rules might provoke him into pulling support for Ukraine or triggering a good old-fashioned transatlantic trade war.

Meta and its ilk had the nerve to moan that enforcing the AI Act could wall off Europe from cutting-edge innovation. It seems that was all it took for Brussels to roll over and beg.

The draft proposes centralising enforcement under a new AI office, probably to make it easier for US lobbyists to know where to send the threats.

For now, EU officials insist they’re still “fully behind the AI Act and its objectives.” Which is true, so long as those objectives don’t upset Washington.