EU Agrees to Scale Back, Delay Parts of AI Act
BRUSSELS — The European Council and European Parliament have agreed to slim down and delay certain provisions of the EU AI Act, according to a Lewis Silkin LLP legal analysis.
The agreement comes as EU lawmakers have faced pressure from the technology industry — including major U.S. AI companies — over the regulatory requirements of the regulation, which is widely considered the world’s most comprehensive AI law.
The EU AI Act, which entered into force in August 2024, established a phased implementation timeline with different categories of requirements taking effect at staggered intervals. The agreement between the Council and Parliament modifies portions of that original timeline, pushing back deadlines for certain compliance obligations and narrowing the scope of some provisions.
Impact on U.S. Companies
The changes carry implications for American technology firms. Companies including OpenAI, Google, Meta, Microsoft and Anthropic — all of which offer AI products and services to European customers — are subject to the regulation’s requirements for their EU-facing operations.
Any delays in implementation timelines or reductions in scope translate directly to extended compliance windows and potentially reduced costs for these companies, many of which have invested heavily in EU AI Act readiness programs.
“The EU AI Act has served as a global regulatory benchmark,” said the Lewis Silkin analysis, noting that changes to the European framework could ripple through legislative efforts in other jurisdictions, including the United States.
Regulatory Context
The EU AI Act uses a risk-based classification system, categorizing AI applications as minimal risk, limited risk, high risk or unacceptable risk. Obligations scale with the risk level, with the most stringent requirements — including conformity assessments, transparency obligations and human oversight mandates — applied to high-risk AI systems.
The regulation also introduced specific rules for general-purpose AI models, including foundation models and large language models, with additional requirements for those deemed to pose “systemic risk.”
The full enforcement of high-risk AI system requirements had been set for Aug. 2, 2026. While certain provisions governing prohibited AI practices took effect in February 2025, the broader compliance framework has been rolling out in phases.
Industry and Legislative Ripple Effects
The decision to reduce portions of the law comes amid a broader policy shift in Europe, where officials have sought to balance innovation-friendly policies with consumer protection. The European Commission under President Ursula von der Leyen has increasingly emphasized competitiveness alongside regulation.
In the United States, the EU AI Act has served as both a template and a reference point for lawmakers crafting domestic AI legislation. More than 40 states have introduced AI-related bills, and several federal proposals have drawn on the EU’s risk-based framework.
The agreement between the Council and Parliament will need to be formally adopted before the changes take legal effect.
