Skip to main contentSkip to main content
Annex III high-risk: prov. 2 Dec 2027 (Omnibus deal, 7 May 2026)
Art. 50 transparency (chatbot/deepfake): still 2 Aug 2026
CRA: Class I/II notifications from Jun 2026 — See CRA tools →

The EU AI Act
Self-Assessment Toolkit

Full legal text, plain-English guides, and compliance tools for the EU AI Act and Cyber Resilience Act — produce a complete evidence package for self-assessment under Art. 43(2) and CRA Annex VIII.

Built for SMEs under 250 staff. Proportionate, plain-English, and no notified body required for most systems.

Self-assessment is the default path for EU AI Act compliance — Art. 43(2)Learn more →

Or start a guided compliance flow if you already know your risk level.

Time until Annex III high-risk enforcement (provisionally 2 Dec 2027 — Omnibus deal)

00

days

00

hrs

00

min

00

sec

Still on track for 2 August 2026, NOT delayed by the Omnibus: Art. 50 transparency duties — chatbot disclosure, deepfake disclosure by deployers, emotion recognition and biometric categorisation notification. Already live: Art. 4 AI literacy (Feb 2025), Art. 5 prohibitions (Feb 2025), Chapter V GPAI obligations (Aug 2025).

Provisional Digital Omnibus deal — high-risk deadlines shifted

On 7 May 2026, the Council of the EU and the European Parliament reached provisional political agreement on the Digital Omnibus on AI. Key shifts:

  • Annex III high-risk obligations: 2 August 2026 → 2 December 2027
  • Annex I product-embedded high-risk: 2 August 2027 → 2 August 2028
  • New Art. 5 prohibition on AI generating child sexual abuse material or non-consensual intimate imagery: 2 December 2026
  • Art. 50(2) machine-readable watermarking ONLY: 2 August 2026 → 2 December 2026
  • Member State AI sandboxes: 2 August 2026 → 2 August 2027

What is NOT delayed by the Omnibus — still live on the original dates:

  • Art. 4 AI literacy — in force since 2 February 2025 (all organisations)
  • Art. 5 existing prohibitions — in force since 2 February 2025
  • Chapter V GPAI obligations — in force since 2 August 2025
  • Art. 50 transparency duties (chatbot disclosure 50(1), emotion recognition / biometric categorisation 50(3), deepfake disclosure 50(4), public-interest disclosure 50(5)) — still 2 August 2026. Only the machine-readable watermarking limb (50(2)) was shifted.

If you deploy a chatbot, a deepfake-generation tool, an emotion-recognition system or a biometric categoriser — your 2 August 2026 deadline is unchanged, even if your underlying AI is "minimal risk".

The Omnibus agreement is provisional. Formal adoption by Council and Parliament is expected before 2 August 2026. If formal adoption fails, the original Regulation (EU) 2024/1689 dates apply across the board.

See regulatory updates for full source citations.

Complete legislation coverage

Full statutory text of every regulation — with plain-English explanations and obligation checklists on every page.

SMEs and corporate teams — using a third-party AI does not exempt you

Using Claude, ChatGPT, Gemini, Copilot, Mistral, Llama, or any third-party AI tool inside your product or business workflow? You are likely a deployer under Art. 26.

The AI vendor (Anthropic, OpenAI, Google, Microsoft, Meta, Mistral, etc.) takes the Art. 16 provider obligations for the underlying model. The duty to operate the system safely in your business context is yours as a deployer under Art. 26 — including:

  • Designating and training a human overseer (Art. 26(2))
  • Ensuring input data is relevant and representative for your context (Art. 26(4))
  • Monitoring system operation and reporting serious incidents to the provider (Art. 26(5)–(6))
  • Informing affected workers, customers or citizens about the AI's role in decisions (Art. 26(7), Art. 26(11))
  • Conducting a Fundamental Rights Impact Assessment if you are a public body or in Annex III 5(b)/(c) (Art. 27)
  • Registering as a deployer in the EU database where you are a public body (Art. 49(2))
  • Maintaining automatically generated logs for at least six months (Art. 26(6))

Compliance templates

Professional documents that satisfy EU AI Act requirements. Buy once, use forever.

See all pricing →

Real-world compliance examples

Worked case studies showing how different AI systems are classified and what compliance looks like in practice.

View all case studies →

Why AIAuditRef

Built specifically for the EU AI Act — not a generic compliance tool.

Full statutory text

Every article, every annex — the actual legal text alongside plain-English explanations on the same page.

Article-linked tools

Every tool shows which article it addresses. Every product template maps to specific compliance obligations.

Enforcement-focused

Built around the real enforcement timeline — including the Digital Omnibus shift of Annex III high-risk to 2 December 2027, while keeping AI literacy, prohibitions, transparency and GPAI obligations on their live dates.

Provider & deployer split

Obligations differ for AI system providers vs deployers. AIAuditRef keeps them clearly separated.

Multi-regulation view

EU AI Act alone isn't enough. CRA, GDPR, Product Liability, and AI Liability all intersect — we cover them all.

Not legal advice

We're a reference and tooling platform. We always remind you to engage qualified legal counsel for your specific situation.

Start your compliance journey today

Free risk classification. Free enforcement timeline. Free compliance checklist. No credit card required.