# EU AI Act dates we track for evidence work

This page is an engineering orientation aid. It is not legal advice. Arviser does not certify compliance or run conformity assessment.

Dates below follow the publicly discussed timeline after the Digital Omnibus adjustments reported in 2026. Confirm every date with your counsel before you rely on it.

| From | What applies | What we can test or document |
| --- | --- | --- |
| 2 Feb 2025 | Prohibited practices (Art. 5) and AI literacy (Art. 4) | Behavioural checks for prohibited patterns in scope; literacy documentation as process evidence |
| 2 Aug 2026 | Transparency (Art. 50): chatbots, synthetic content, deepfakes | Whether the system identifies as AI; labelling of generated content under your configuration |
| 2 Dec 2026 | Machine-readable marking for generative systems already on the market | Technical marking expectations for media outputs where in scope |
| 2 Dec 2027 | Annex III high-risk obligations (employment, credit, insurance, education, essential services) | Logging, human oversight, data-quality notes, and documentation packs for counsel — not a certificate |
| 2 Aug 2028 | High-risk AI built into regulated products (Annex I) | Boundary notes when an LLM component sits inside a regulated product |

## How Arviser uses this

Optional **EU AI Act evidence add-on** on Readiness or Audit: technical tests and documentation inputs your counsel can use. Never “AI Act compliant” as our outcome.

**Next:** [LLM Quality Audit](/llm-quality-audit) · [Evidence Pack](/evidence-pack) · [Request scope](/#scope)
