Skip to content
Innopulse Consulting
EU AI Act·● Pillar article

The EU AI Act in 2026: what has applied since August and what follows until 2028 — the SME guide after the Digital Omnibus

The AI Act after the Digital Omnibus: timeline to 2028, roles, risk classes, AI literacy, transparency, penalties and what SMEs should do now.

Leutrim Miftaraj
Leutrim Miftaraj
Founder & CEO
·Updated ·10 min read

The EU AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024 and applies in stages. In July 2026 the EU amended it with the Digital Omnibus, Regulation (EU) 2026/1744: it entered into force on 27 July 2026, postpones the high-risk obligations and softens the AI literacy duty. Much material online still reflects the old position. This guide describes the state as of October 2026 and what it means for small and medium-sized companies.

The timeline after the Omnibus

  • 2 February 2025: bans on certain AI practices and the AI literacy duty under Art. 4 apply.
  • 2 August 2025: obligations for providers of general-purpose AI (GPAI) models and the governance structures apply.
  • 2 August 2026: most of the Regulation applies, including the transparency obligations under Art. 50; national authorities supervise and enforce.
  • 2 December 2026: end of the transition for machine-readable marking under Art. 50(2) for generative systems on the market before 2 August 2026.
  • 2 December 2027: obligations for high-risk systems under Annex III.
  • 2 August 2028: obligations for high-risk systems in regulated products under Annex I.

Postponing the high-risk obligations does not mean nothing applies until 2027. Bans, AI literacy, transparency and GPAI obligations already apply, and since August 2026 there are authorities enforcing them.

The role comes first

The AI Act allocates obligations by role. The provider is whoever develops an AI system, or has it developed, and places it on the market under its own name; the deployer is whoever uses an AI system professionally under its own authority. A company using ChatGPT, Copilot or an AI tool in HR is a deployer. A software house building AI into its product is the provider of that product. The role is determined per system, not per company.

The risk classes

Prohibited practices under Art. 5, such as social scoring, manipulative techniques or certain forms of emotion recognition in the workplace and education, have been banned since February 2025; the overview is in the glossary under prohibited AI practices. High-risk systems are the use cases listed in Annex III, such as recruitment, creditworthiness assessment or education, plus AI as a safety component of regulated products. Certain systems carry transparency obligations regardless of risk class. Everything else is minimally regulated. Separately, providers of GPAI models have their own obligations.

What concretely applies since August 2026

For most SMEs, two obligations matter most today. First, AI literacy under Art. 4: providers and deployers must take measures to support the AI literacy of their staff and others acting on their behalf. The Omnibus turned this from an obligation of result into an obligation of effort; no specific level per person must be guaranteed, but measures must be taken. Details in our article on Article 4 after the Omnibus. Second, transparency under Art. 50: chatbots must disclose that an AI is answering, deepfakes and certain AI texts on matters of public interest must be labelled, and providers of generative systems must mark content in machine-readable form. The Commission adopted guidelines on this in July 2026, and a code of practice on marking has been available since June 2026.

High-risk: prepare rather than wait

Anyone providing or deploying an Annex III system has until December 2027. The obligations are extensive: risk management, data governance, technical documentation, logging, human oversight, accuracy and robustness, conformity assessment, registration and post-market monitoring. Certain deployers must produce a fundamental rights impact assessment. Many of these requirements cannot be retrofitted in a few weeks because they concern architecture and processes. Our article on high-risk obligations describes them in detail.

Penalties

Art. 99 tiers the fines. Breaches of the bans can be fined up to EUR 35 million or 7 percent of worldwide annual turnover, breaches of other obligations, such as for high-risk systems or transparency, up to EUR 15 million or 3 percent, and incorrect information to authorities up to EUR 7.5 million or 1 percent. For SMEs the lower of the two amounts applies. The AI literacy duty is not separately quantified in this catalogue; member states set the penalties for it.

What SMEs should do now

1. Build an AI inventory: which systems are in use, for what, in which role.
2. Classify each system: prohibited, high-risk, subject to transparency, or minimal. Our high-risk classification checklist guides through this.
3. Plan, run and document role-based AI literacy measures.
4. Implement transparency: AI notice for chatbots, labelling of published AI content, clarification with tool vendors on machine-readable marking.
5. Review contracts with AI vendors: data use, training, documentation, support with obligations.
6. For high-risk candidates, draw up a roadmap to December 2027.

And Switzerland?

The AI Act also applies to Swiss companies when they place AI systems on the EU market or when their systems’ output is used in the EU. Switzerland itself does not adopt the Regulation but takes its own path. What that means in practice is described in our article The EU AI Act and Switzerland.

Support

For classifying individual systems we built our own tool, AI Risk Check, which performs classification under Art. 6 with legal citations and produces audit-ready reports. For companies needing a structured overall view we offer an AI Act gap assessment. Legal assessment of individual cases remains the job of specialised law firms; our strength is technical and organisational implementation.

About the author
Leutrim Miftaraj
Leutrim Miftaraj
Founder & CEO · Innopulse Consulting

Founder of Innopulse Consulting, IT project lead and business analyst. Mandates incl. EDA, MeteoSwiss, FSO, SERI, FINMA, Swisscom and BLS. MSc FFHS, PMP, IREB CPRE. Author of “Identity Over Discipline”.

Topics
eu ai act 2026ai act timelinedigital omnibus ai actai act smeai act obligations companiesai act switzerland
Working on something similar?

Let's talk.

If this article maps to a problem you're actively working on, send us a short description — we'll respond with a practical next step.