- As of July 27, 2026, Article 4 of the EU AI Act requires measures rather than guaranteed results. This change was introduced by Regulation (EU) 2026/1744, known as the Digital Omnibus on AI.
- No one has to guarantee a certain level of competence among individual employees anymore.
- The European Commission does not require a specific curriculum or certificate. An internal list of training courses is sufficient as proof.
- National market surveillance authorities have been monitoring compliance with the AI Act since August 2, 2026.
- The requirements for high-risk systems under Annex III will take effect on December 2, 2027, and those under Annex I on August 2, 2028.
- The AI Act does not apply directly in Switzerland. It applies through the EU market.
On July 27, 2026, the EU scaled back its own requirements regarding AI literacy. The obligation remains, but its nature has changed. Anyone who concludes that training is now a secondary concern is misinterpreting the change.
What exactly was changed
Article 4 has been in effect since February 2, 2025. At the time, it was brief and worded in a rather demanding manner. The Digital Omnibus Act has replaced it.
Through July 26, 2026
Obligation to Succeed
„to ensure, to the best of their ability, that their staff has a sufficient level of AI expertise"
A result was needed.
Since July 27, 2026
Duty to Use Best Efforts
„Measures to Support the Development of AI Competence," with no guarantee for any individual
We owe it to ourselves to make a serious effort.
Legally speaking, this is the difference between an obligation to achieve a specific result and an obligation to make a good-faith effort. In the past, a company was obligated to achieve a specific result. Today, it is obligated to make a sincere effort.
In practice, this shifts the focus of the question of evidence. The question was: Can you demonstrate that your team has AI expertise? It now reads: Can you show that you've done something about it?
An Overview of the Deadlines
- 02.02.2025 Article 4 shall enter into force in effect
- 27.07.2026 Article 4 has been watered down in effect
- 02.08.2026 Market Surveillance Begins in effect
- 02.12.2027 High Risk, Appendix III open
- 02.08.2028 High Risk, Appendix I open
What Counts as a Measure
The European Commission has published a separate Q&A on this topic. It lists four steps as the minimum requirements:
- Build a general understanding of AI within the organization.
- Clarify our role. Are we a provider or an operator?
- Assess the risks associated with the systems in use.
- Derive specific measures based on this, tailored to the employees’ prior knowledge, the task at hand, and the context in which they work.
According to the Commission, an internal list of training sessions and initiatives is sufficient as proof. The Q&A states verbatim: „There is no need for a certificate."
That's a sentence that providers of certification courses are reluctant to quote. We do it anyway, because it's true and because any compliance department can find it within two minutes.
Does this apply to Swiss companies?
The AI Act is EU law and does not apply directly in Switzerland. However, it has extraterritorial effect. As soon as an AI system is placed on the market in the EU or its output is used in the EU, the regulation applies, regardless of where the company is headquartered.
In practical terms, this means that anyone who has EU customers, operates a subsidiary in the EU, or is a supplier in an EU supply chain falls within the scope of the regulation. For many Swiss SMEs, this is the norm.
Added to this is an effect that has very little to do with it, as is only right. Major clients assess AI expertise in requests for proposals and supplier audits long before a regulatory agency comes to inspect.
Why a Recognized Certificate Still Matters
This is precisely the point that the slowdown brings into focus. Compliance was never the best argument for AI training; it was just the loudest one.
That leaves three reasons, and all three are more stable than a legal situation that has changed twice in 18 months.
First: The certificate belongs to the person
An internal training record serves as documentation for the employer. An internationally recognized credential appears on your resume and remains valid even if you change jobs.
Second: A tested curriculum covers what never comes up when you're just experimenting.
Anyone who teaches themselves about AI becomes good at prompting. But there are some things that are reliably left out of the picture: data quality and bias, the limitations of the models, data protection, and the question of who in the company is accountable for an AI system. These are the issues that can end up costing a lot.
Third: An audit creates a binding obligation
Someone who knows there will be 40 questions at the end will approach the test differently than someone who is sure they have a participation sheet.
Why we're addressing this topic in the first place
We train business analysts, product owners, requirements engineers, software architects, project managers, testers, and software developers. Since 2023, we’ve also been welcoming beginners, starting with ChatGPT 101 and most recently with the AI Masterclass.
Our largest course, the IREB CPRE Foundation, has always been a certification course. The exam is administered by IREB, not by us. It was different for the AI courses. In those courses, everyone received a certificate of participation from Obvious Works, along with the information that we are certified trainers from swissAI are. Both were honestly earned, and neither was subject to an external review.
This began to bother us more and more. A certificate issued by the course provider itself primarily serves as proof of attendance.
We are closing this gap with the A4Q AI Foundation, in partnership with Shajan Academy, an accredited A4Q training provider. The exam is administered by iSQI. A4Q is the certification brand, and iSQI administers the exam.
What the Course Covers
Two days, 9:00 a.m. to 5:00 p.m. each day, with a maximum of 12 participants. Official A4Q curriculum covering eight subject areas.
| Day | Area | Key Topics |
|---|---|---|
| 1 | Introduction to AI | History, Symbolic and Data-Driven AI, Common Misconceptions |
| 1 | Technology Overview | Machine learning, neural networks, NLP, computer vision, training data, data quality and bias, AI agents |
| 1 | Revenue and Efficiency | Customer Behavior, Pricing, Supply Chains, Marketing and Service Automation, Tool Selection |
| 1 | Prompt Engineering | Creating Effective Prompts, Types of Prompts, Limitations, Iterative Improvement |
| 2 | AI Assistance Systems | Areas of Application, Integration, Dependencies, Measurement of Efficiency Gains |
| 2 | Regulation, Ethics, Safety | Legal Framework, EU AI Act, Data Protection, Fairness, Human Oversight, Manipulative Attacks |
| 2 | Change Management | Engage employees, address resistance, and develop roles and competencies |
| 2 | AI Roadmap | Four Dimensions, Use Case Prioritization, Pilot Implementation, Governance, Success Measurement |
Table with horizontal scrolling
The exam is administered online with remote proctoring, in German or English. The course costs CHF 2,500, with a 10 percent early-bird discount. Meals are included; for online sessions, they are provided in the form of a voucher. The course also includes an online refresher and a second chance to take the exam.
Current Events
A curriculum is adopted as of a specific date, but the technology continues to evolve. This is the case for every certification course, regardless of the provider.
We teach the official A4Q curriculum in its entirety and without modification. It serves as the basis for the exam, and that is what we prepare students for. In addition, our two trainers—who work with these systems on a daily basis—bring their expertise to the table: current examples, the latest tools available at the time of the course, and insights into which of these are applicable to your own organization.
The curriculum provides the structure. The relevance comes from the classroom.
What it doesn't replace
Honesty is part of the topic, which is why I've included this section.
- No Guarantee of Compliance
- The AI Act does not require a certificate. Anyone who claims that a course will make a company AI Act-compliant is selling a sense of security they cannot deliver.
- No AI Governance
- Two days build expertise. They do not replace any policy, any inventory of the systems in use, or any areas of responsibility.
- No technical training
- Anyone who trains models or builds AI features into software needs a different format.
- No expertise
- No one becomes an AI expert in just two days. The program is called “Foundation,” and that’s exactly what it means.
Frequently Asked Questions
Does our company now have to purchase training courses because of Article 4?
No. Article 4 requires measures that are appropriate for the team’s prior knowledge and operational context. For a team that uses AI only to a limited extent, an internal training session may be sufficient. An external course is one possible measure among several.
What happens in the event of violations?
Article 4 does not provide for any specific penalty. Penalties are governed by the national laws of the Member States and must be proportionate. However, market surveillance authorities may take AI competence into account as part of other assessments.
Does Article 4 also apply to management and external parties?
Yes. This requirement applies to employees and other individuals who work with AI systems on behalf of the company. This includes freelancers and service providers.
We're in Switzerland. Does this affect us?
Not directly. But generally speaking, through EU customers, EU subsidiaries, and supply chains, yes.
How does the A4Q AI Foundation differ from the AI4RE Micro-Credential?
A4Q AI Foundation is designed for all roles within an organization and provides broad coverage of AI, from technology to regulation to roadmaps. AI4RE is tailored to requirements engineering.
Is prior knowledge required?
No. The course is designed as an introductory course.
Next Step
All dates and registration are handled through our ticket shop. For teams, the course is also available as an in-house format; simply send a brief inquiry via the course page.
Michael (Mr. Miroboard) Mey
Michael is a trainer who not only impresses with his knowledge, but also with his passion.
About Michaels Trainer profile
To his LinkedIn profile
DE 
EN