EU AI Act — High-Risk & GPAI Enforcement Begins

What this countdown tracks

Countdown to 2 August 2026, when the bulk of the EU AI Act's substantive obligations become enforceable. From that date, providers of high-risk AI systems and general-purpose AI (GPAI) models face the full supervisory and penalty regime, including administrative fines of up to €15M or 3% of worldwide annual turnover for non-compliance with GPAI duties.

Background

Regulation (EU) 2024/1689 — the AI Act — entered into force on 1 August 2024. Its obligations apply in staggered phases set out in Article 113: prohibitions on unacceptable-risk practices started on 2 February 2025; GPAI model obligations for newly placed models took effect on 2 August 2025; and the main enforcement milestone for high-risk systems and supporting governance kicks in on 2 August 2026, two years after entry into force.

The Act classifies AI uses into four risk tiers. High-risk systems — covering areas such as biometric identification, critical infrastructure, education, employment, essential services, law enforcement, migration, and justice — must meet detailed requirements on risk management, data governance, logging, transparency, human oversight, accuracy, and robustness. GPAI model providers, including those behind frontier models, must publish training-data summaries, respect EU copyright law, and, for systemic-risk models above the 10^25 FLOP training-compute threshold, conduct model evaluations, track serious incidents, and report to the European AI Office. Fines scale from €7.5M / 1% for supplying incorrect information up to €35M / 7% for prohibited-practice violations. The general-purpose AI Code of Practice, coordinated by the AI Office with plenary input from model providers, civil society, and academics, was finalised in mid-2025 and functions as a presumption-of-compliance tool until harmonised standards catch up. Major signatories include Anthropic, Google, Microsoft, and OpenAI; Meta and xAI publicly declined the initial draft, flagging contested provisions on training-data disclosure.

Why the date matters

2 August 2026 is the statutory switch-on for the core supervisory architecture: national competent authorities designated, notified bodies operational, the European AI Office empowered to exercise GPAI oversight, and the conformity-assessment and post-market monitoring duties binding for most high-risk providers. It is the first date on which the Commission can open a full non-compliance procedure against GPAI providers under Article 101. Fines landing from that day onward will set precedents for model documentation, copyright compliance, and systemic-risk evaluation standards.

What to watch for

• Publication of the final GPAI Code of Practice and its take-up by major model providers.
• First enforcement actions by the AI Office against systemic-risk model providers.
• Member-state designation of national market surveillance authorities.
• Harmonised standards from CEN-CENELEC covering risk management and data governance.
• Guidance on copyright opt-outs and text-and-data-mining training datasets.
• Interaction with GDPR supervisors over training-data processing.
• Early post-market incident reports filed under Article 73.

Related events to track

The EU AI Act Phase 2 countdown on 2 August 2027 extends compliance to legacy GPAI and embedded high-risk systems. Enforcement overlaps with the DMA first review countdown in May 2026, where scope may expand to AI. The GDPR 10-year review countdown will examine data-protection interactions with AI training.

FAQ

When exactly is EU AI Act enforcement? 2 August 2026 for high-risk systems and the main GPAI enforcement regime, under Article 113.

Is EU AI Act enforcement confirmed or expected? Confirmed by the text of Regulation (EU) 2024/1689 as published in the Official Journal.

Who is responsible for EU AI Act enforcement? The European AI Office within DG CONNECT handles GPAI; national competent authorities handle high-risk systems; the European Artificial Intelligence Board coordinates.

Where can I read the official announcement? The consolidated timeline is at https://artificialintelligenceact.eu/implementation-timeline/ with the full legal text on EUR-Lex.