Swedish prime minister urges pause to EU AI Act rollout

Swedish Prime Minister Ulf Kristersson has broken ranks with his European counterparts to demand a pause in the implementation of the EU's Artificial Intelligence Act, telling parliament in Stockholm on Monday that the regulation is "confusing" and risks leaving European companies unable to comply because the technical standards it depends on do not yet exist.

Kristersson said he would raise the issue at this week's European Council meeting in Brussels, making him the first sitting head of government to call publicly for a halt to the rollout. Officials in the Czech Republic and Poland have previously signalled openness to a delay, but no other leader has put the demand on the table at a summit.

The standards gap at the heart of the problem

The AI Act entered into force in August 2024 after years of negotiation between the Council, Parliament and Commission. It is being phased in over an 18-month period, with the first prohibitions on unacceptable-risk systems taking effect in February 2025 and the bulk of obligations for high-risk systems due by August 2026. But the legislation was written to rely on harmonised technical standards, developed by European standardisation bodies under a mandate from the Commission, to give companies a practical route to compliance on requirements such as cybersecurity, data governance, transparency and human oversight.

Those standards are not ready. The European Committee for Standardisation (CEN) and the European Committee for Electrotechnical Standardisation (CENELEC) are still working on them, and industry groups have warned that the timeline is too tight. Without the standards, companies face legal uncertainty: they must meet the Act's essential requirements but have no presumption of conformity to rely on. Kristersson described this as a regulatory failure. "An example of confusing EU regulations is the fact that the so-called AI Act is to come into force without there being common standards," he told the Riksdag's EU committee.

Commission acknowledges the difficulty

Henna Virkkunen, the Commission's executive vice-president for tech sovereignty, security and democracy, has not dismissed the idea of a pause. In recent weeks she told MEPs that if the necessary guidance and standards are not ready in time, a delay to certain parts of the Act is possible. That marks a shift from the Commission's earlier insistence that the timetable was fixed. Virkkunen's portfolio includes the digital simplification package expected before the end of 2025, which is intended to cut red tape across EU technology legislation.

The Commission's willingness to entertain a pause reflects growing pressure from member states and industry. BusinessEurope, the confederation of European employers, has argued that the standards process is running years behind the legislative calendar. German and French industry associations have echoed that warning, particularly for small and medium-sized enterprises that lack the legal and engineering resources to interpret the Act's requirements without the standards.

Political momentum builds in the Council

Kristersson's intervention gives political weight to a discussion that has until now been confined to technical working groups. The Czech presidency of the Council's working party on telecommunications and information society raised the possibility of a delay last autumn, and Polish officials have since made similar noises in private. But the European Council agenda for this week's meeting does not list the AI Act as a dedicated item, meaning Kristersson will have to raise it under 'any other business' or during a working lunch.

Diplomats in Brussels say several capitals are sympathetic but wary of reopening a legislative file that took three years to close. France and Germany, the two largest economies, have not publicly backed a pause. Paris in particular sees the Act as a competitive advantage for European AI firms that can market themselves as compliant with the world's first comprehensive AI law. Berlin is more cautious but has emphasised the need for legal certainty rather than a formal delay.

Parliament's conservative wing breaks cover

The most striking political signal came from the European Parliament. Arba Kokalari, a Swedish MEP sitting with the European People's Party, the Parliament's largest group, issued a statement backing Kristersson. "If standards are not ready in time, we should stop the clock for certain parts of the AI Act and give companies more time," she said. Kokalari also urged the Commission to include the AI Act in its digital simplification package, a move that would open the legislation to amendment less than a year after it entered into force.

That position is a departure for the EPP, which led the Parliament's negotiating team on the Act and has defended the final text. Kokalari's intervention suggests the centre-right is hearing complaints from business constituents in Sweden, Germany and the Benelux countries. The Parliament's other main groups, Socialists & Democrats, Renew Europe, Greens/EFA, have not echoed the call. A senior S&D aide said the group would resist any move that "undermines the credibility of the legislative process".

What a pause would mean in practice

A formal pause would require a legislative act, either a targeted amendment to the AI Act's transitional provisions or a separate regulation extending the deadlines. Both would need approval from the Council and Parliament, a process that typically takes six to twelve months. The Commission could also issue a recommendation to national market surveillance authorities to defer enforcement of specific obligations, but that would not remove the legal obligation itself.

Industry lawyers say the most pragmatic route would be for the Commission to adopt an implementing act extending the date by which harmonised standards must be available, coupled with a declaration that companies acting in good faith will not be penalised during the interim. That would avoid reopening the legislative text while giving standard-setters more time. The Commission has used similar mechanisms for the Medical Devices Regulation and the Machinery Regulation when standards lagged behind.

Competitiveness arguments cut both ways

Kristersson warned that continuing the rollout without standards could lead to "Europe falling behind technologically or specific applications not being available on the European market." That argument, that over-regulation strangles innovation, has been the centrepiece of the Commission's own competitiveness narrative under President Ursula von der Leyen's second term. The digital simplification package is explicitly framed as a response to Mario Draghi's report on EU competitiveness, which identified regulatory burden as a drag on productivity.

But the counter-argument is that the AI Act's credibility depends on its predictability. Investors in European AI startups have cited the regulatory framework as a reason to back European firms over US or Chinese rivals, precisely because the rules are clear and enforced. A pause, especially one driven by political pressure rather than technical necessity, could signal that the EU's regulatory commitments are negotiable. That risk was highlighted by the European Consumer Organisation (BEUC), which warned that delaying obligations on high-risk systems such as biometric identification or credit scoring would leave consumers unprotected.

The digital simplification package as a vehicle

Kokalari's proposal to fold the AI Act into the digital simplification package is the most concrete procedural path on the table. The package, due before the end of 2025, is expected to target reporting obligations, data access rules and cross-border compliance costs across a dozen digital laws. Adding the AI Act would allow the Commission to propose targeted deadline extensions alongside broader simplification measures, potentially insulating the move from accusations of a U-turn.

The Commission has not confirmed whether the AI Act will be included. A spokesperson said the package's scope is still being finalised and that any changes to the AI Act would respect the "balance struck between innovation and fundamental rights". That language suggests the Commission will resist substantive amendments to the risk classifications or prohibited practices, but may be open to transitional adjustments.