
EU agrees to simplify the AI Act. The transparency clock now runs faster.
EU Council and Parliament reached political agreement on May 7 to simplify AI Act rules. The headline: a tightened 3-month transparency deadline for synthetic content, plus carveouts for SMEs and small mid-caps.
EU Council and Parliament reached political agreement on May 7 to amend several AI Act provisions [Council press release]. The package, part of the "Omnibus VII" simplification agenda, changes both deadlines and obligations.
── What shipped ──
Three concrete changes:
Faster transparency deadline. The grace period for providers to implement transparency solutions for AI-generated content drops from 6 months to 3 months. The new deadline is December 2, 2026.
Slower regulatory sandboxes. The deadline for member states to establish AI regulatory sandboxes is pushed to August 2, 2027 — a year later than the original timeline.
Wider SME carveouts. Regulatory exemptions previously available only to SMEs are extended to small mid-caps (SMCs). Sensitive personal data may be processed for bias detection and mitigation in a wider set of cases.
The AI Act itself entered into force August 1, 2024 and becomes fully applicable August 2, 2026 [Implementation timeline]. Prohibited practices and AI literacy obligations have applied since February 2025. Governance rules and obligations for general-purpose AI models have applied since August 2025.
── Why it matters ──
For US-headquartered companies serving EU customers, the December 2, 2026 transparency deadline is the line item. If your product generates synthetic media — images, audio, text-as-output, voice — it must be machine-detectable as AI-generated by that date. Three months is a tight runway for legal-product-engineering coordination, especially for vendors that have been treating EU compliance as a 2027 problem.
The SME / SMC carveout is more pragmatic than political. Bias mitigation requires using sensitive personal data, which the original Act made awkward. The amendment removes that contradiction.
The sandbox postponement is the quiet story. Regulatory sandboxes are how startups get permission to test high-risk AI systems under supervision. Pushing the deadline by a year means national regulators are not on track to provide that permission infrastructure on time.
── Editor's take ──
The simplification narrative will be debated by every think tank in Brussels for the next month. For shipping teams, ignore the politics and put one calendar entry: December 2, 2026, AI-content provenance must ship to EU users. The rest is noise that lawyers handle.
// newsletter_offline · provider_not_configured