AI Integration Guide

EU AI Act Deadlines 2026–2028: The Current Timeline

The EU AI Act's high-risk deadlines were pushed back in 2026 — here is the current, officially confirmed timeline through 2028 and which date applies to you.

25 Jul 2026 ·5 min read ·Pranoti Kshirsagar
EU AI ActAI complianceAI regulationAI governance

The EU AI Act’s original 2 August 2026 deadline for high-risk AI obligations no longer applies. The European Parliament approved a simplification package on 16 June 2026 (423 votes in favor, 57 against, 174 abstentions) that pushed several key dates back, in some cases by well over a year. This guide gives the current, officially confirmed timeline, sourced directly from European Parliament press releases. It points to which date applies to your situation.

Source note: Every date below is quoted or confirmed directly from official European Parliament press releases (europarl.europa.eu). Where a figure could not be confirmed against an official EU source, this guide leaves it out rather than including it as a guess. This page is dated 15 July 2026. Legislative dates can still shift with further procedural steps (Official Journal publication, entry into force). Check the lastmod date above against the official sources linked here before you rely on this for a compliance deadline.

What you need before starting

  • No technical setup. This is a reference guide to confirm which deadline applies to your organization.
  • If you are unsure whether your tool or model falls under the AI Act at all, start with the AI system definition test before you check dates.

Why the dates changed

The Commission’s original 2024 timeline set 2 August 2026 as the date most high-risk AI obligations would apply. Parliament’s simplification package changed that. The European Parliament’s own press release states that the co-legislators intended to adopt the package “before 2 August 2026, the start date for current rules on high-risk systems.” This confirms the original deadline was the trigger for the reform, not a date still in effect.

The dates that still stand

Two dates from the original AI Act timeline are unaffected by the 2026 simplification package:

  • 2 February 2025 — Prohibited practices (Article 5), AI-literacy duties, and the AI system definition (Chapters I and II) began applying.
  • 2 August 2025 — Rules for general-purpose AI models, governance, notifying authorities, confidentiality, and penalties began applying.

EU AI Act deadlines that changed in 2026

According to the European Parliament’s press release on the simplification deal:

  • Stand-alone high-risk AI systems (including biometrics, critical infrastructure, education, employment, law enforcement, and border management use cases): obligations now apply from 2 December 2027, not 2 August 2026.
  • High-risk AI systems used as safety components covered by EU sectoral product-safety legislation (for example, machinery, medical devices): obligations now apply from 2 August 2028, not 2 August 2027.
  • Watermarking obligations on AI-generated content: postponed to 2 December 2026. The Commission’s original proposal had set 2 February 2027, so this date is earlier than what the Commission itself proposed, though later than the original AI Act date.

This watermarking deadline is a distinct, provider-facing obligation, separate from the Article 50(4) content-disclosure duty covered in How to add AI disclosures that comply with the EU AI Act. The two are easy to conflate but are not the same requirement.

What is new in the package

Alongside the deadline changes, the package adds a new prohibition. It bans AI systems that generate non-consensual sexual or intimate content, or child sexual abuse material (CSAM), often called “nudifier” apps. Companies must bring existing systems into compliance with this ban by 2 December 2026.

EU AI Act deadlines 2026–2028 at a glance

DateWhat applies
2 February 2025Prohibited practices, AI-literacy duties, AI system definition
2 August 2025GPAI model rules, governance, penalties
16 June 2026Parliament approves the simplification package
2 December 2026Watermarking obligations for AI-generated content; nudifier/CSAM ban compliance deadline
2 December 2027Stand-alone high-risk AI system obligations
2 August 2028High-risk AI systems embedded in EU sectoral-regulated products

Which date actually affects you?

  • If you provide a general-purpose AI model, the 2 August 2025 obligations already apply to you. See GPAI Code of Practice: Provider vs. Deployer to confirm whether you are a provider at all.
  • If you deploy a stand-alone high-risk AI system (for example, in recruitment, education, or law enforcement contexts), you now have until 2 December 2027, not 2026.
  • If your AI system is a safety component in an already-regulated product (medical devices, machinery), your deadline is 2 August 2028.
  • If you publish AI-generated content and need to add machine-readable watermarking, your deadline moved to 2 December 2026.
  • If your product could generate non-consensual intimate imagery or CSAM, you must comply with the new prohibition by 2 December 2026, regardless of your risk category.

Troubleshooting

“I thought I had until August 2026 for high-risk compliance.” That was the original deadline. The June 2026 simplification package pushed it to 2 December 2027 for stand-alone high-risk systems, or 2 August 2028 if your system is embedded in an already-regulated product.

“Is this simplification package still just a proposal?” No. Parliament formally approved it on 16 June 2026 with a clear majority (423-57-174). This guide does not confirm the exact Official Journal publication or entry-into-force date. Treat these as the adopted legislative dates rather than assume every procedural step is complete.

“Does the watermarking deadline apply to me if I just write about AI, not build it?” No. Watermarking is a provider-facing technical obligation on the AI system itself. If you publish AI-generated text or images for public-interest topics, the relevant duty is the Article 50(4) disclosure obligation covered in the linked guide above, not this one.

What you can do now

You can now state, with a source you can point to, which AI Act deadline applies to your specific situation. Some deadlines moved later (high-risk systems) while others moved earlier than the Commission’s own proposal (watermarking).

This guide accompanies EU AI Act Article 4: What AI Literacy Requirements Mean for European Research Institutes on The Science Talk. Note that the Article 4 AI-literacy duty (applicable since 2 February 2025) was changed by the 2026 simplification package covered here: the Omnibus simplified the company-level requirement and shifted the leading role to the Commission and Member States.


Want more guides like this? Browse all AI Guides or get in touch →

Found this useful? Share it or explore more guides.