27JUL

Brussels fixes AI Act job deadline

2 min read

10:02UTC

The Council of the European Union gave final adoption to the Digital Omnibus on 29 June, fixing the AI Act's high-risk employment deadline at 2 December 2027.

← Back to SAP freezes R&D headcount as others deny AI Jump to analysis ↓

EconomicDeveloping

Key takeaway

The AI Act's employment deadline is now law, but the worker's right to an explanation carries no enforcement.

The Council of the European Union (EU) gave final adoption to the Digital Omnibus, the package that finalises the AI Act's workplace rules, on 29 June. The European Parliament had formally endorsed it at its 16 June plenary, though the rules do not bind until the text appears in the Official Journal, the EU's legal gazette, entering force three days later. ¹

The deadline for high-risk employment uses of AI, meaning decisions on hiring, monitoring or dismissal, is NOW fixed in law at 2 December 2027. That gives employers about 18 months before the rules bite, while European workers face the displacement NOW and the safeguards later. The worker AI-literacy duty, which would have given staff a right to an explanation of the systems used on them, stays non-binding.

Parliament had already voted 423 to 57 to strip enforceability from that literacy duty ; the Council's adoption turns the awaited formal votes into settled law. BusinessEurope backed the softer wording, arguing a duty to 'ensure' worker comprehension was impossible to meet. The European Trade Union Confederation (ETUC) called the shift from 'ensure' to 'support' a collapse of the legal threshold rather than a drafting tidy-up, because a duty to 'support' gives a worker no cause of action.

Deep Analysis

In plain English

The EU has a big AI law called the AI Act. New rules about how AI can be used to hire, monitor and fire workers were meant to include a duty for employers to make sure staff understand how the AI systems affect them. On 29 June, EU governments gave final approval to a package called the Digital Omnibus that waters that duty down: instead of employers having to 'ensure' workers understand AI, they now only have to 'support' understanding. That might sound like a small wording change, but it matters: 'ensure' means a worker could complain if their employer failed to deliver, while 'support' gives the employer no specific target to fail. The tougher AI workplace rules themselves are still coming, just not until 2 December 2027.

Deep Analysis

Root Causes

The 'ensure' to 'support' downgrade followed the same Council unanimity dynamic that produced the Minimum Wage Directive's 'promote' language: any binding worker-facing duty requiring active enforcement infrastructure needs member states to fund inspectors and appeals processes, a cost several governments were unwilling to commit to inside the broader Digital Omnibus simplification push.

BusinessEurope's lobbying succeeded because the Digital Omnibus was framed from the outset as a 'simplification' exercise aimed at reducing compliance burden, which structurally favours softening any obligation industry can characterise as burdensome, regardless of its individual merits.

What could happen next?

Consequence
Workers in EU member states now have no EU-wide legal claim if an employer fails to explain how a workplace AI system affects them, since the 'support' duty carries no enforcement mechanism comparable to GDPR's breach-notification regime.
Immediate · Assessed
Risk
France and Germany, both of which have domestic labour codes that already exceed EU minimums, are likely to legislate stricter AI-literacy obligations at national level, fragmenting the single market for AI-employment compliance rather than harmonising it.
Medium term · Suggested
Meaning
The 2 December 2027 high-risk employment deadline is now the only binding date left in the package, making it the single concrete test of whether the AI Act delivers any enforceable workplace protection at all.
Long term · Assessed

Source Landscape

This story draws on centre-left-leaning sources from United States

United States

Primary parallel: The EU's 2022 Adequate Minimum Wage Directive replaced 'guarantee' with 'promote' for collective-bargaining coverage targets specifically because Council unanimity requirements could not accommodate ETUC's preferred binding language, the same drafting technique NOW applied to AI literacy.

Counter-parallel: GDPR's Article 32 uses the same binding 'ensure' verb for data security and has generated billions of euros in enforcement fines since 2018, showing an 'ensure' duty is enforceable when paired with a dedicated regulator, breach-notification duty and audit trail, infrastructure the AI-literacy provision was never given.

Immediate impact: employers avoid the compliance cost of building AI-literacy training programmes with audit trails, a cost Bruegel's earlier drafting analysis put at several hundred euros per employee for large multinational deployers.

Structurally, the softened duty removes the main lever that would have let EU workers force disclosure of how workplace AI systems are used on them, leaving the December 2027 high-risk deadline as the only binding date left in the package unless France or Germany legislate a stricter domestic floor.

Consensus view: Bruegel's technology-policy team assess that shifting the employer AI-literacy duty from 'ensure' to 'support' removes any statutory cause of action: a worker who is not trained on a workplace AI system NOW has no legal claim against their employer, only a duty the employer can satisfy by gesture rather than outcome.

Counter-view: BusinessEurope counters that the binding 'ensure' standard was impossible to audit at scale, since AI-literacy training cannot be independently verified the way a fire-safety or GDPR training log can, and argues 'support' simply matches the law to what a regulator can actually inspect.

Key tension: Whether an unenforceable duty represents a defensible retreat from an unauditable standard, or whether it hands employers a permanent excuse never to build the auditing infrastructure regulators would need to enforce anything stronger.

Sources:Bloomberg

Mentions:Bruegel →Brussels →Parliament →Official Journal →European Union →Council of the European Union →European Parliament →Digital Omnibus →EU AI Act →BusinessEurope →European Trade Union Confederation →GDPR →

First Reported In

Update #15 · Oracle names AI in its own annual report

Bloomberg· 1 Jul 2026

Read original →

Causes and effects

This Event

Brussels fixes AI Act job deadline

European workers face displacement now while the binding safeguards arrive in December 2027, and the right to an AI-use explanation stays non-binding.

Led to

EU waters down worker AI rights

The Council's final adoption follows directly from Parliament's 16 June 423-57 vote stripping enforceability from the literacy duty.

Occurred 16 Jun 2026

Read story →

EU AI rules await their final votes

The Council's adoption turns the formal votes previously awaited into settled law.

Occurred 20 May 2026

Read story →

California binds AI hiring, EU defers

The EU Council's final adoption of the Digital Omnibus enabled the Annex III deferral now taking effect.

Occurred 1 Jul 2026

Read story →

Germany names the AI Act's enforcers

The EU Council's 29 June adoption of the Digital Omnibus set the deadline Germany's Bundesrat is now enforcing nationally.

Occurred 10 Jul 2026

Read story →

A Spanish court opens the algorithm

The Digital Omnibus's final adoption set the 2 December 2027 compliance deadline the Bundesrat and the Commission's guidelines now work against.

Occurred 25 Jul 2026

Read story →

Different Perspectives

European Commission

The European Commission's draft Annex III guidelines, closed for comment on 23 July, treat algorithmic scoring in recruitment, pay and termination as high-risk regardless of whether a human signs off, echoing Spain's Audiencia Nacional ruling 101/2026 on concealed scheduling algorithms. Brussels is shifting the fight from counting AI job losses to assigning legal liability for the tools themselves.

Office for National Statistics

The Office for National Statistics recorded UK vacancies rising to 712,000 on 21 July, the first quarterly increase this beat has tracked, with payrolled employment down 85,000 on the year against May's 210,000 fall. The bulletin names no AI cause anywhere, and that is the point: nothing in the release confirms the displacement story it gets cited to support.

Christian Klein, SAP

Christian Klein told investors on 23 July that SAP's research headcount will not grow for twelve months because AI agents and their token costs are absorbing the work, not because SAP is cutting jobs. He frames it as commercial arithmetic: the cost of AI-assisted coding tokens plus the salaries specialist AI hires command, not people being replaced by machines.

Betsey Stevenson, University of Michigan

Betsey Stevenson argued that the 187,000 jobless-claims reading describes a market that hires little and fires little, not one AI is emptying. She said the real damage hides in eligibility rules and suppressed job postings, not in the headline layoff counts employers keep denying.

Comisiones Obreras, UGT and Concentrix's A Coruña works committee

Comisiones Obreras, UGT and Concentrix's A Coruña works committee blamed Microsoft's push toward AI self-service for the 80 redundancies unions signed off on 22 July, not unavoidable business cause. A second Coruña procedure covering 80 more jobs runs to a 31 August deadline, and the unions want the state, not the employer, setting the pace of AI-driven cuts.

Stanford's 'We Must Act Now' signatories

More than 200 academics, including 16 Nobel laureates, published a 13 July letter warning of AI-driven labour disruption, citing Daron Acemoglu's NBER estimate that AI's total factor productivity gain stays under 0.66% over ten years. The letter's own cited economics sit well below Goldman Sachs Research's 1.5-percentage-point estimate published the same week.