EU AI Act deadlines: what is due when in 2026-2027

The EU AI Act is not a single deadline, it is a phased schedule. For CTOs and product teams, missing the wrong date can mean fines of up to EUR 15 million or 3% of global revenue. This guide maps every material deadline through 2027 so you can build a compliant roadmap without surprises.

What the EU AI Act requires

The Act entered force on August 1, 2024, and has been rolling out in stages. February 2, 2025 was the deadline to cease any prohibited AI practices under Article 5 (unacceptable risk). August 2, 2026 is the next critical date: obligations for General Purpose AI model providers under Articles 51 to 56 take effect, and transparency requirements for AI-generated content under Article 50 also apply from this date. The deadline for high-risk AI systems listed in Annex III, which covers areas including hiring, credit, education, and critical infrastructure, is December 2, 2027. This deadline was extended from August 2, 2026 under the Digital Omnibus Act passed in May 2026. High-risk systems covered by existing EU product safety directives (Annex II) faced obligations from August 2, 2026.

What this means for your business

The practical consequence of the phased schedule is that different parts of your product may have different compliance horizons. If your platform uses a large language model and exposes AI-generated outputs to users, Article 50 labelling obligations land in August 2026, giving you roughly two months from today. If your product includes a CV screening, credit scoring, or student assessment feature, you have until December 2027 for Annex III obligations, but technical documentation (Article 11) and risk management systems (Article 9) take months to build properly, so starting now is not early. GPAI model providers, including any startup that has built and distributes a foundation model, face the August 2026 deadline and must register in the EU database.

Steps to get compliant

1. Audit every AI feature in your product and assign it to a risk tier and a deadline by running it against the Annex III categories and Article 6 criteria. 2. Treat August 2, 2026 as your immediate sprint target: implement Article 50 transparency labelling for any AI-generated text, images, audio, or video in your product. 3. For any Annex III high-risk feature, begin building your Article 9 risk management system and Article 11 technical documentation now, targeting a complete compliance package by mid-2027 to allow time for audit and remediation. 4. Set calendar reviews every quarter to catch any changes from EU implementing acts, delegated regulations, or harmonised standards that could affect your obligations.

Free EU AI Act risk assessment

Not sure where your AI system stands? The free ActComply risk screener classifies your system in under 5 minutes. No sign-up required.