The 2 August 2026 compliance trap: why the EU AI Act already affects your daily business

The “Digital Omnibus” has been in force since 27 July 2026: with it, the EU postpones the strict obligations for high-risk systems until December 2027 and softens individual rules – such as the AI literacy obligation. The reaction in the mid-market is understandable – but risky: “So we still have time, we can put the topic off until next year.”

This is exactly where the error lies. Thomas Fischer, AI expert at PASSION4IT, has been watching the developments for months and warns: the relief affects only part of the EU AI Act. The obligations for completely ordinary, everyday AI use – that is, what your teams have long been doing with ChatGPT, Copilot or Midjourney – went live on 2 August 2026.

“Anyone sitting this out isn’t risking Article 4 fines – there are none. They’re risking their team blowing the labelling obligation. And that one has been backed by fines since August.” — Thomas Fischer, AI expert at PASSION4IT

The EU AI Act is therefore no longer an abstract Brussels construct that only concerns tech giants. It explicitly puts the users of AI under obligation, too. Three points belong on every management team’s radar now.

1. AI literacy obligation (Art. 4) – still applies, with a new rationale

This obligation has applied since February 2025 and remains in place after the Omnibus – though adjusted: companies must foster the AI literacy of their employees, not guarantee a specific level. It carries no fine of its own. Why the topic still belongs at the top of the list: for deployers of high-risk AI, qualification remains binding, documented training is your central piece of exonerating evidence in a dispute – and without trained employees, the fine-backed labelling obligation (point 2) is practically impossible to meet. Those who don’t recognise when they’re producing AI content don’t label it.

2. Transparency obligation (Art. 50) – since 2 August 2026

AI-generated content must be clearly labelled as such. Service chatbots, AI texts in marketing, voice clones: it must be recognisable to users that they are interacting with a machine. For existing systems there is a technical grace period until December 2026 (for example for watermarks in images), but the fundamental disclosure obligation already applies. And unlike Art. 4, fines are provided for here: up to EUR 15 million or 3 percent of global annual turnover; the Omnibus halves the rates for SMEs. In Germany, this has been supervised since 2 August 2026 by the Bundesnetzagentur as the central AI supervisory authority.

3. Supply-chain obligation (Art. 23 & 24) – since 2 August 2026

Anyone who imports AI tools or passes them on becomes part of the AI supply chain and needs contractual assurances from their software vendors that their models meet the European requirements.

The good news: you don’t have to manage this alone

From our consulting practice at PASSION4IT we see it every day: most companies don’t underestimate the technology, but the burden of proof. AI has long been in use – but hardly anyone can document who was trained, which rules apply and how content is labelled. This is exactly where we step in. For each of the three obligations, we have the right answer.

For the literacy obligation: the AI driver’s licence from our Academy. With the AI driver’s licence we bring your teams up to the required competence level in a hands-on way – and issue a certificate that gives you proof under Art. 4 in black and white. That way you don’t just meet the obligation, you can demonstrate it at any time. Exactly what counts in a dispute.

For the transparency and supply-chain obligations: our AI consulting. Together with you we draw up a legally sound AI usage policy, clarify labelling obligations and check your AI supply chain for gaps. And the best part: this consulting can be subsidised through the BAFA funding programme – but only until the end of 2026. Here, too, acting quickly pays off twice over.

For the overview: the AI readiness check. Don’t know where your company stands? Our AI readiness check shows you in just a few minutes which of the three obligations you already meet – and where gaps remain.

The 2nd of August 2026 has arrived – the baseline obligations apply. Those who start now have their training documented, their policies in place and their funding secured before the first review comes.

Take the first step: start the AI readiness check now – free of charge and in just a few minutes. Or book a call with our team directly.

Further resources

• AI driver’s licence: EU AI Act Art. 4 mandatory training for mid-market companies — how to meet the Art. 4 literacy obligation in a documented and demonstrable way.
• AI readiness check for companies in the DACH region 2026 — stocktake and concrete next steps before August 2026.

Frequently Asked Questions

Which EU AI Act obligations apply since 2 August 2026?

Since 2 August 2026, the baseline obligations for everyday AI use have applied: the fine-backed transparency obligation under Art. 50 (labelling AI-generated content) and the supply-chain obligations under Art. 23 and 24. The AI literacy obligation under Art. 4 has already applied since February 2025. Since 2 August 2026, the Bundesnetzagentur has also been the central AI supervisory authority in Germany.

Wasn’t the EU AI Act postponed?

Only in part. The “Digital Omnibus” entered into force on 27 July 2026. It postpones the strict obligations for high-risk systems until December 2027 and softens the literacy obligation under Art. 4 from “ensure” to “foster”. The baseline obligations for normal, everyday AI use — with ChatGPT, Copilot or Midjourney, for instance — are unaffected and have applied since 2 August 2026.

What does the AI literacy obligation under Art. 4 require?

Companies must demonstrably foster the AI literacy of their employees — with documented training, attendance records and internal policies. Art. 4 provides for no fine of its own; the documentation is, however, the central piece of exonerating evidence in liability questions and the foundation for meeting the labelling obligation under Art. 50. The AI driver’s licence from the PASSION4IT Academy provides this proof with a certificate.

How must AI-generated content be labelled?

Under Art. 50, users must be able to recognize that they are interacting with a machine or seeing AI-generated content — this covers service chatbots, AI-written marketing copy and voice clones. For the technical implementation in existing systems (e.g. watermarks in images) there is a grace period until December 2026; the basic labelling obligation has applied since 2 August 2026 and is backed by fines of up to EUR 15 million or 3 percent of global annual turnover (reduced by 50 percent for SMEs).

Who is affected by the supply-chain obligations under Art. 23 and 24?

Any company that imports or passes on AI tools becomes part of the AI supply chain and needs contractual assurances from software vendors that their models meet the European requirements.

How do I become compliant now?

In three steps: the AI driver’s licence for the literacy obligation, a legally sound AI usage policy including labelling and supply-chain checks for Art. 50/23/24, and — up front — the AI readiness check to establish where you stand. The consulting can be subsidized through BAFA funding, but only until the end of 2026.

This article is for information purposes and does not constitute legal advice.