The European Union reached a second major AI Act agreement today. On May 7, 2026, the Council presidency and European Parliament negotiators struck a provisional deal on a dedicated streamlining proposal - separate from the Digital Omnibus on AI agreed in April - that brings targeted changes to the EU’s AI compliance framework.

For businesses operating in Europe, there are three things in this agreement that deserve attention: who now qualifies for simplified compliance, a new hard deadline for AI-generated content labelling, and a shift in how high-risk AI timelines work.

SME Relief Extended to Medium-Sized Companies

The most significant expansion in today’s agreement is that the simplified compliance requirements previously available only to small and micro-enterprises now extend to small mid-caps. The definition covers companies with up to 500 employees.

In practice, this means a larger group of businesses can now use simplified technical documentation requirements for general-purpose AI systems. If you’re a company under 500 employees that has been preparing for the full documentation burden, this agreement gives you more room to work with.

This matters because the technical documentation requirements for general-purpose AI are substantial. Reducing that load for companies under 500 employees significantly broadens who can deploy AI at reasonable compliance cost.

AI-Generated Content: A Shorter Window Than Expected

Here is the provision that will catch some companies off guard. The original EU AI Act gave providers six months to implement transparency measures for AI-generated content - things like labelling AI-created images, audio, and video. Today’s agreement cuts that window in half.

The grace period is now three months, and the hard deadline is December 2, 2026. Any provider generating content using AI - images, synthetic audio, deepfakes, AI-written video narration - needs transparency mechanisms in place by that date.

This applies broadly. It includes tools that generate synthetic media for marketing, social content, advertising, training materials, or customer communications. If your business or your vendors are producing AI-generated content at scale, you need to confirm that labelling and disclosure mechanisms will be live by December 2.

High-Risk AI: Standards-Contingent Timeline

For high-risk AI systems, today’s agreement shifts the compliance trigger from a fixed calendar date to a conditional one. Rules will apply once the European Commission confirms that the required harmonised standards and testing tools are in place, with a maximum extension of 16 months from the original deadline.

This provides some breathing room for companies deploying AI in regulated sectors - healthcare, critical infrastructure, education, recruitment - where certification requirements depend on technical standards that are still being developed. The practical implication is that the deadline is no longer fully predictable. Businesses should monitor Commission announcements rather than planning to a fixed date.

Other Changes

The agreement also postpones the deadline for national governments to establish AI regulatory sandboxes from August 2026 to August 2, 2027. For companies that were hoping to use national sandbox programs to test high-risk AI under controlled conditions before full compliance requirements kick in, that infrastructure will arrive later than expected.

The deal additionally reinforces the powers of the EU AI Office and takes steps to reduce fragmentation in how different member states implement governance. This should, over time, reduce the risk of companies having to navigate twelve different national interpretations of the same underlying regulation.

What Needs to Happen Before December 2

The most urgent action item in today’s agreement is the AI-generated content deadline. Companies have approximately seven months to:

1. Audit all AI tools and workflows that generate content visible to end users
2. Determine which outputs fall under EU transparency obligations
3. Implement labelling, metadata tagging, or watermarking as appropriate
4. Verify that third-party vendors and platforms you rely on have their own compliance in place

The April agreement already pushed back several high-risk AI deadlines. This agreement adds a new countdown that applies to a much wider set of businesses. Any company producing marketing content, social media, customer-facing materials, or product images using AI tools is potentially in scope.

What This Means for Business

Two EU AI Act agreements in two weeks reflects the pace at which European policymakers are trying to calibrate a framework that was written before the current generation of AI tools existed. The direction is toward simplification for smaller companies and a conditional approach for complex compliance areas, while tightening specific consumer-facing requirements like content labelling.

The net message for most businesses is: if you’re under 500 employees, your compliance burden for general-purpose AI just got lighter. If you use AI to generate content, your labelling deadline just got stricter. And if you’re deploying high-risk AI, you’re working to a moving target tied to when the Commission publishes the required standards.

This is one of the more actively shifting regulatory environments in business right now. Staying current is not optional.

---

Enterprise DNA helps businesses navigate what AI regulations actually mean for their operations - not just what they say in legal text. If you’re working through EU AI Act compliance alongside your AI deployment plans, talk to us.