The European Commission is steering the amendment drive for the AI Act, with delegated annex updates under ordinary legislative procedure; DG Connect drafts texts with input from the AI Board and GPAI codes, aiming for finalization in early 2026, while the Act aligns with the Digital Services Act and Data Act and carries projected EU-wide enforcement costs of 4–10 billion euros annually.

The EU plans to finalize positions on proposed amendments by April 2026, concentrating on enforcement, high-risk systems, and general-purpose AI models, with coordinated input from the Commission, the Council, and the Parliament.

France, Germany, and the Netherlands are leading voices on balancing innovation with safety, and insights from 2025 pilots and feedback from over 200 organizations in late 2025 dialogues are shaping the proposed changes.

Coordination structures such as the AI Technical Working Party and COREPER, along with Commission impact assessments, guide revisions; active sandboxes across multiple states and 2025–2026 pilots inform decision making.

Implementation timeline shows prohibitions already in effect, high-risk system requirements kicking in from August 2026, obligations for general-purpose AI starting August 2025, and late-2025 public consultations on Codes of Practice for foundational models.

EU influence on global standards is advancing, with references in Japan and South Korea, consideration of US 2026 bills, and alignment with G7 goals, as other jurisdictions pursue related governance and cross-border harmonization.

Stakeholder input comes from BusinessEurope, Parliament committees, rights groups, national parliaments, and SMEs via EESC opinions, supplemented by sandbox deployment data from 15 states.

Amendments target Annex III high-risk lists (including biometric identification and education tools), potential adjustments to systemic risk thresholds for general-purpose models, and penalties up to 35 million euros or 7% of turnover, with ongoing discussions on enforcement harmonization and standardized penalties.

The AI Act, in force since August 2024, uses a risk-based approach, bans certain practices, requires risk assessments for high-risk uses (e.g., hiring, medical devices), and mandates transparency for generative AI, with national authorities enforcing rules aided by the European AI Office.