BRUSSELS — The European Commission is moving to soften the EU AI Act by delaying key compliance deadlines and pushing back when regulators can start levying major penalties until 2027, a shift officials frame as “simplification” but critics call a capitulation to Big Tech lobbying and rising transatlantic pressure, Dec. 24, 2025.
The rollback is packaged inside a broader “Digital Omnibus” effort that aims to trim red tape across the bloc’s digital rulebook. But the most politically charged piece is the EU AI Act: rules for some “high-risk” systems would slide later than expected, giving companies more time to complete conformity assessments and documentation before enforcement bite dates arrive.
EU AI Act: why Brussels is hitting the brakes
The Commission’s proposal would push certain high-risk obligations from 2026 into late 2027, according to Reuters reporting on the draft. The delay is not just technical. It lands after months of warnings from U.S. and European firms that the compliance runway is too short, and amid louder U.S. criticism that EU tech enforcement amounts to discriminatory regulation.
In public messaging, the Commission argues that the EU AI Act remains intact and that staged implementation still moves forward — just more slowly in areas where standards, guidance and enforcement capacity are not fully ready. The EU’s own timeline notes that the law rolls out in phases and does not fully apply across all categories until 2027; the Commission-hosted service desk lists Aug. 2, 2027, as a key milestone for high-risk AI in certain regulated products. See implementation timeline.
Lobbying, “competitiveness,” and a transatlantic squeeze
The politics are harder to simplify than the paperwork. The proposal arrives as European leaders fret that the continent is falling behind the U.S. and China on advanced AI. It also follows sustained industry campaigning for more time and fewer obligations, plus warnings that overlapping EU rules could chill investment. Euronews described the plan as a retreat that would give companies more time to adapt, while prompting pushback from some lawmakers and experts.
Digital rights groups counter that the EU AI Act was designed precisely to put guardrails around sensitive uses — from hiring and credit scoring to biometric systems — and that postponing penalties weakens deterrence. The Guardian reported that parts of the package were criticized as a broader rollback of digital protections, with AI compliance delays among the flashpoints.
Continuity: this fight has been building for years
The EU AI Act has been a political battleground since well before it became law. The European Parliament adopted the landmark regulation in March 2024 after a December 2023 political agreement, pitching it as a rights-forward framework for trustworthy AI. (Background: European Parliament press release, March 2024.)
Even before final passage, law firms and industry groups flagged looming implementation pain points — including standards, governance and the compliance load for “high-risk” systems. (Background: White & Case analysis, December 2023.)
And as the first milestones approached, companies and lobby groups warned the EU not to “stop the clock” — a debate that now looks newly revived. In July 2025, Reuters reported the Commission publicly rejected calls to pause the rollout, insisting there would be no blanket delay.
Now, the Commission’s new approach signals that enforcement posture is shifting, even if the EU AI Act text stays on the books. The proposal still needs scrutiny from EU member states and the European Parliament — setting up a high-stakes fight over whether Europe’s first-of-its-kind AI law keeps its teeth, or trades them for time.
