Civil Society Warns EU AI Act Weakening as Commission Bypasses Standard Procedures
Over 40 organisations challenge Digital Omnibus amendments, citing missing impact assessments and lack of public consultation on AI Act changes.
Civil Society Raises Alarm Over EU AI Act Amendment Process
More than 40 civil society organisations, led by ARTICLE 19, have formally flagged serious concerns about the European Commission’s approach to amending the AI Act through the so-called Digital Omnibus package. The core issue: the Commission appears to have sidestepped standard legislative procedures, including mandatory impact assessments and public consultation.
What’s Being Changed
The amendments under discussion would delay high-risk AI system compliance timelines, with the European Parliament adopting a position that pushes implementation deadlines to December 2027 for biometric and critical infrastructure systems, and August 2028 for systems covered by existing EU sectoral legislation.
While the Parliament’s position includes specific, fixed dates (which civil society acknowledges adds some predictability), the process by which these changes were developed has raised red flags about transparency and accountability.
Why This Matters
The AI Act was supposed to represent the gold standard in democratic regulatory design—one of the few pieces of technology legislation developed with extensive stakeholder consultation. The current approach to amendments appears to undermine that foundation.
For Ireland and EU member states preparing enforcement frameworks, procedural shortcuts at the EU level create additional uncertainty. Ireland’s nascent AI Office, due to launch by August 2026, will inherit ambiguity about which compliance deadlines are final and which might shift again.
Civil society organisations argue that bypassing impact assessments means regulators, builders, and affected communities lack evidence-based justification for the changes. This is particularly concerning for vulnerable populations potentially affected by high-risk AI systems in education, employment, and law enforcement.
The Irish Perspective
Ireland’s distributed enforcement model—with 15 sectoral competent authorities coordinating through a new AI Office—depends on clear, stable rules. Frequent amendments developed without proper procedure add compliance costs and create confusion about which authority has responsibility for what.
For Irish tech companies, the message is stark: delays to implementation deadlines sound appealing, but they come wrapped in procedural uncertainty that could make future amendments even more unpredictable.
What Remains Unclear
- Whether the Commission will acknowledge these procedural concerns or proceed with Digital Omnibus amendments as drafted
- How Ireland’s AI Office will interpret conflicting deadlines if amendments proceed without proper impact assessment
- Whether future AI Act amendments will follow similar expedited procedures
- What remedies civil society has if the Commission continues to bypass standard legislative consultation
The coming weeks will reveal whether EU institutions take these concerns seriously, or whether political pressure to move fast overrides procedural safeguards that protect both regulated entities and the public.
Source: ARTICLE 19