EU AI Act Update — Key Developments
1. Political agreement reached on the "Digital Omnibus" AI changes
EU lawmakers and Member States reached a provisional agreement on amendments to the AI Act aimed at simplifying compliance and delaying some obligations. The agreement still requires formal approval before it becomes final law.
Most important proposed changes
- High-risk AI obligations would move from 2 August 2026 to 2 December 2027.
- Some machinery and sector-regulated products may be excluded from overlapping AI Act requirements.
- National AI sandbox deadlines may shift from 2026 to 2027.
- New bans would target AI systems generating non-consensual explicit imagery ("nudifier apps").
- Mandatory watermarking obligations for AI-generated content are expected to strengthen.
2. Important clarification: current deadlines still legally apply
Although political agreement has now been reached, organisations should not assume the delay is fully enacted yet.
Multiple legal and governance analyses stress that:
- the original AI Act remains in force;
- August 2026 obligations technically still apply until the amendments are formally adopted and published.
Current official timeline
According to the European Commission:
- prohibited AI practices and AI literacy obligations already apply;
- GPAI model obligations already apply;
- high-risk AI system obligations are currently scheduled for 2 August 2026;
- some embedded product AI systems transition until 2027.
3. Enforcement and governance trend
The broader EU direction remains clear:
- stronger AI governance,
- stronger platform accountability,
- tighter interoperability and competition controls,
- increasing oversight of foundation models and AI ecosystems.
The European Commission is also continuing work on:
- AI Office oversight,
- regulatory sandboxes,
- GPAI supervision,
- coordination between AI Act, GDPR, DMA, NIS2, CRA, and sector-specific rules.
4. What this means for organisations now
Recommended actions
Continue preparing as if August 2026 remains the operational target date.
Priority areas:
- AI inventory and classification
- supplier and model governance
- human oversight procedures
- AI literacy programmes
- documentation and audit trails
- transparency and labeling practices
- vendor due diligence
- data governance and cybersecurity controls
For web and digital teams
Particular attention should be paid to:
- AI-generated website content,
- AI search experiences,
- chatbots and virtual assistants,
- automated personalization,
- accessibility tooling,
- synthetic media and image generation,
- AI-assisted customer service workflows.
5. Website-ready takeaway
The EU is moving toward a softer and more business-friendly implementation of the AI Act, but organisations should continue preparing for compliance now. While lawmakers have agreed in principle to delay some high-risk AI obligations until late 2027, the current legal deadlines remain in force until the amendments are formally adopted. Businesses should treat this period as a governance preparation window rather than a pause in compliance activity.