21SEP

EU AI rules await their final votes

3 min read

16:45UTC

The EU Digital Omnibus still needs formal Council and Parliament votes before an August adoption target, the last step that would lock in a softened worker-facing rule, or reopen it if the timetable slips.

← Back to AI closes the hiring door, not the exit Jump to analysis ↓

EconomicDeveloping

Key takeaway

Formal Council and Parliament votes against an August target will decide whether and when the rules bind.

The EU Digital Omnibus, the Brussels package bundling amendments to the AI Act, Data Act and Cybersecurity Act, still awaited formal Council and Parliament votes as of late May 2026. ¹ A provisional deal struck on 7 May had already pared back the employer duty on worker AI literacy , but a provisional deal is not statute. Until both institutions ratify the text, the softened wording is a negotiated position, not a rule firms must follow.

The package targets adoption in August. That timetable matters because the votes are the last juncture at which the text can still move. Parliament can reopen clauses it conceded in trilogue, and a Council delay past the summer recess would push the AI Act's high-risk employment obligations further out without the worker-facing literacy layer Parliament had built around them. The provisional deal itself emerged only after the 28 April trilogue collapsed and reconvened in mid-May , so the negotiation has already shown it can stall.

European workers feel the difference in timing, not wording. If the votes land on schedule, the weakened obligations become law in August and run alongside the AI Act's December 2027 high-risk deadline. If they slip, European workers spend longer with no statutory AI-transparency right in force while the displacement count keeps climbing. The substance was decided in trilogue; what these votes decide is whether, and when, it binds.

Deep Analysis

In plain English

The EU Digital Omnibus is a package of EU law amendments. It updates several existing laws at once, including the EU AI Act, which set rules for how AI can be used in workplaces across all 27 EU member states. One of the AI Act's original provisions required employers to ensure their workers actually understand the AI systems being used in their jobs. This obligation was designed to prevent situations where AI makes decisions about a worker's performance or eligibility for promotion without the worker being able to understand or challenge those decisions. During negotiations in May 2026, that binding obligation was weakened. Instead of ensuring workers understand AI, employers now only need to take measures to support that understanding, a phrase with no enforcement mechanism and no penalty for non-compliance. The negotiators from the European Parliament, the EU Council (representing member state governments), and the European Commission reached a provisional agreement on 7 May. A provisional agreement requires formal votes in both institutions before it takes legal effect. For it to become law, both the full European Parliament and the full Council must formally vote to adopt it. As of late May 2026, those votes have not yet happened. The adoption target is August 2026. Once it passes, the high-risk employment AI obligations take effect in December 2027. Until then, no employer AI-literacy duty exists in EU law, weakened or otherwise.

Deep Analysis

Root Causes

The Omnibus weakening of the AI-literacy obligation traces to two concurrent pressures. First, the European Parliament's March 2026 vote to delay high-risk employment deadlines reflected the same productivity-focused MEP coalition that then negotiated the Omnibus literacy weakening.

Second, the 28 April trilogue collapse created deadline pressure on all parties: with the August adoption target set by the Cypriot Presidency, a deal requiring the Parliament to accept Council positions on employer obligations was the path of least resistance.

The formal Council and Parliament votes remain the last procedural gate. Until the Council and Parliament votes conclude, employers face no enforceable obligation under the weakened text. Any MEP or member state that objects to specific provisions can still raise procedural objections in the formal vote, though a provisional deal that both institutions' lead negotiators signed is rarely rejected at the formal stage.

What could happen next?

Precedent
If the Omnibus passes as provisionally agreed, the EU sets a legal precedent that employer AI-literacy obligations in employment are advisory rather than mandatory, weakening worker leverage in individual AI-system disputes.
Risk
The December 2027 high-risk employment AI deadline applies without the literacy mechanism that would allow workers to verify compliance; national labour inspectorates inherit technically complex AI audit obligations they were not resourced to perform.

Source Landscape

This story draws on neutral-leaning sources

Primary parallel: The EU's 1994 European Works Council Directive required multinational companies operating across member states to establish transnational worker consultation bodies. The final text of the 1994 Directive was substantially weaker than the 1990 Commission draft on information and consultation rights, after a negotiation between Parliament (pushing for binding rights) and Council (pushing for employer flexibility) that mirrors the 2026 Omnibus dynamic.

The 1994 compromise text was later strengthened in the 2009 recast Directive, suggesting that weakened worker-facing EU employment obligations have historically been revisited once the political coalition that drove the weakening shifts.

Counter-parallel: The EU's 2002 Information and Consultation Directive was also weakened in trilogue before adoption, but the weakened text was never subsequently strengthened and remains the floor for worker information rights across the EU today. The counter-precedent is that weakened EU worker-protection directives are not always revisited: some become permanent floors.

Consensus view: Reinhilde Veugelers (Bruegel, Brussels) and Philipp Lausberg (European Policy Centre) have argued that the Digital Omnibus negotiations reflect a broader EU competitiveness-versus-protection tension that the 2025 Draghi report on European competitiveness quantified: EU AI adoption by firms runs at roughly 40% of the US rate, partly attributed to compliance overhead.

The Omnibus weakening of employer obligations is consistent with the Draghi report's recommendation to reduce regulatory burden on AI deployment, but it comes at the direct cost of the worker-facing provisions the European Parliament had prioritised.

Counter-view: Agnieszka Jankowiak (CEPS, Centre for European Policy Studies) and Lora Borissova have documented that the employer AI-literacy obligation's weakening removes the only procedural mechanism through which EU workers could establish whether an AI system affecting their employment was operating as described by their employer.

Without that baseline literacy requirement, the AI Act's discrimination and transparency provisions for high-risk employment systems are formally enforceable but practically unverifiable by the affected workers. This creates a structural enforcement gap that the December 2027 employment deadline inherits.

Key tension: Whether weakening the AI-literacy duty meaningfully accelerates EU AI adoption and productivity (the Draghi competitiveness argument) or simply transfers the compliance monitoring burden from employers to underfunded national labour inspectorates that lack the technical capacity to audit AI systems.

Sources:AI Laws By State

Mentions:Digital Omnibus →Bruegel →European Parliament →Brussels →EU AI Act →Parliament →European Commission →

First Reported In

Update #10 · Rival studies split on AI's hit to jobs

AI Laws By State· 24 May 2026

Read original →

Causes and effects

Caused by

Brussels drops binding AI literacy duty in Digital Omnibus

The 7 May provisional deal that weakened the employer AI-literacy obligation is the substance now awaiting formal Council and Parliament votes ahead of the August adoption target.

Occurred 7 May 2026

Read story →

EU Digital Omnibus second trilogue collapses

The 28 April trilogue collapse and 13 May resumption produced the provisional deal whose formal votes remain the final procedural step.

Occurred 28 Apr 2026

Read story →

Brussels fixes AI Act job deadline

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

Occurred 29 Jun 2026

Read story →

This Event

EU AI rules await their final votes

The weakened EU worker protections are not yet law; the formal votes are the last point they can change.

Led to

Brussels dilutes the one worker AI right

The Digital Omnibus was already awaiting formal votes after the May provisional deal; the June committee approval is the direct progression.

Occurred 2 Jun 2026

Read story →

Different Perspectives

Salesforce, Synopsys and TD Bank Group

Salesforce, Synopsys and TD Bank Group each filed quarterly disclosures in late August booking restructuring charges, or none at all, without naming AI as a cause. Their silence matters because Challenger's tracker shows AI as a stated reason fell to fourth place in August even as the year-to-date AI-cut total still leads at 116,175.

Singapore, South Korea, Taiwan and Indonesia

Singapore launched its Skills and Workforce Development Agency on 16 September, giving citizens six months of free premium AI tools, while South Korea ring-fenced its AI tax windfall in a new Future Response Fund. Taiwan kept funding its AI build past NT$190bn and Indonesia rewired vocational training around AI literacy, betting state-built skills beat a market-led adjustment.

ver.di, CGT Fonction Publique and CCOO

Germany's ver.di banked a 3.3% pay rise on 1 September and opened talks on a Tarifvertrag Transformation covering dismissal bans and reskilling, while France's CGT rejected Paris's AI negotiating timetable the same week. Spain's CCOO went further on 21 September, proposing to tax companies by the jobs they generate rather than wait for the next bargaining round.

BIS General Manager and Federal Reserve governors

The BIS's General Manager said on 10 September that AI displacement remains limited, even as the BIS's own survey found nearly 80% of firms plan to automate roles. Two Federal Reserve governors made the same point in July, arguing the labour-market data does not yet show a mass-firing event.

Bank of Canada, ONS and ECB

The Bank of Canada found the job-finding gap between AI-exposed and unexposed occupations widened from 2.2 to 13.9 percentage points since 2015-19, while separations barely moved. That framing, a hiring freeze rather than a firing wave, is echoed by the ECB's finding that euro-area AI use hit 52% of workers in 2026, concentrated among the university-educated.

Office for National Statistics

Deferred its Transformed Labour Force Survey beyond November 2027 and disclosed a May 2026 telephone-collection failure. The ONS carries no AI-attribution layer at all, so Britain sits outside this month's cohort of measuring states by its own admission.