EU Parliament formally endorses AI Omnibus

On 26 March, the European Parliament formally adopted its position on the Digital Omnibus on AI, paving the way for interinstitutional negotiations with the European Commission and the Council.

The vote marks a key step in the EU’s effort to recalibrate its landmark AI Act, as co-legislators face mounting pressure to reach an agreement before core provisions begin to apply in August 2026.

While broadly endorsing the Commission’s mandate to partially de-regulate the AI rulebook, the Parliament’s position introduces a number of tweaks, particularly on timelines for compliance and on anti-nudification provisions.

Clarity on application timelines

The most consequential aspect of the Parliament’s position concerns application timelines. In contrast to the Commission’s original proposal, which kept timelines for high-risk AI obligations open and tied to supporting standards, MEPs have introduced fixed deadlines.

High-risk systems specifically listed in the regulation, including those involving biometrics and those used in critical infrastructure, education, employment, essential services, law enforcement, justice and border management, would face obligations from 2 December 2027.

Systems covered by EU sectoral safety and market surveillance legislation would follow on 2 August 2028, while providers of AI-generated content would have until 2 November 2026 to comply with watermarking requirements.

The shift to fixed dates reflects a broader frustration from industry that open-ended timelines make compliance planning impossible.

Slashing red-tape

In other areas, the Parliament’s position mirrors the Commission’s drive for simplification and reducing red tape for business.

Regulatory relief measures previously available only to SMEs would be extended to small mid-cap enterprises, acknowledging that companies scaling beyond SME thresholds should not face an abrupt compliance cliff.

MEPs also proposed that AI Act obligations should be less stringent for products already regulated under sectoral laws. This will cover sectors including medical devices, radio equipment and toy safety, among others and aims to avoid duplicating compliance requirements across overlapping frameworks.

These provisions reflect recommendations from the 2024 Draghi report, as well as strong industry pressure, to avoid overlapping regulations and reduce the compliance burden for small to mid-tier operators in order to boost innovation in Europe.

Privacy and anti-nudification protection 

Parliament also used its position to go beyond simplification, introducing a new ban on AI “nudifier” systems, namely tools that generate or manipulate sexually explicit or intimate images of identifiable real persons without their consent.

The ban would not apply to systems with effective technical safeguards preventing such misuse.

The Parliament’s text also proposed an ex-ante mechanism for the Commission to adopt delegated acts to ban harmful AI systems that arise due to future technological developments.

Speaking during the plenary vote, co-rapporteur Michael McNamara (IE, Renew) hailed the nudification ban as a signal that non-consensual intimate imagery will no longer be “a price to be paid” for public life in the digital age.

This proposal comes amidst a proliferation of nudification content on platforms such as X in recent months, underscoring the balance lawmakers seek to maintain as they reconcile pro-innovation policies with the need to protect human dignity and fundamental rights.

MEPs also pushed back on Commission proposals to relax data processing conditions for AI bias detection, insisting instead that personal data may only be processed for that purpose where strictly necessary. This standard aligns closer with GDPR principles, indicating certain limits to the Parliament’s embrace of the Commission’s deregulatory path.

Full-steam ahead?

Trilogue negotiations are now moving quickly into motion. The first technical trilogue already takes place on 27 March–one day after the Parliament’s vote–with co-legislators hoping to secure a final agreement by the second trilogue on 28 April. 

This ambitious schedule reflects political pressure to finalise the package before key AI Act provisions are set to legally take effect on 2 August 2026. Failure to reach a deal could produce a legal blackhole over the Act’s applicability.

Though largely welcomed by industry, the Parliament’s position has faced opposition by privacy activists and civil society groups. European Digital Rights (EDRi) warned that the text weakens the oversight power of fundamental rights bodies, among other crucial safeguards.

Other criticism posits that by ripping up measures that were legislated only two to three years ago, the Parliament, in tandem with the other European insitutions, is giving mixed signals to businesses, citizens and other jurisdictions, in turn undermining the EU’s credibility as a global actor. How these tensions will shape the outcome remains uncertain, as co-legislators enter detailed trilogue negotiations against an ambitious legislative timetable.