In this article7 sections
  1. What Article 3 Says About Deployers
  2. Article 4: The Obligation That Is Already In Force
  3. What This Looks Like in a 50-Person Company
  4. The Gap Most Companies Have Right Now
  5. Where GDPR and the AI Act Overlap
  6. Three Steps to Take This Month
  7. References

Does the EU AI Act apply to ChatGPT? Yes. If your employees use ChatGPT, Copilot, or a similar general-purpose AI tool for their work, your company qualifies as a deployer under EU AI Act Article 3(4). That single fact brings you under Article 4, which has applied since 2 February 2025.

Most SME leaders assume the AI Act is a problem for Big Tech. It is not. The regulation explicitly covers the organisations that use AI systems, not just the companies that build them. And the obligation that reaches almost every business, Article 4, is already live.

What Article 3 Says About Deployers

The Act defines a deployer as "a natural or legal person, public authority, agency or other body using an AI system under its authority", except for personal, non-professional use. The definition is intentionally broad. When an employee pastes a customer complaint into ChatGPT to draft a response, your company is using that AI system. When your HR team screens CVs with an AI-assisted tool, your company is the deployer. The provider is OpenAI or Microsoft. You are the deployer.

Size does not grant an exemption from Article 4. A 25-person logistics firm using AI-assisted route planning has the same AI literacy obligation as a 2,000-person corporation, although what counts as reasonable measures will differ. The heavier deployer duties in Article 26 are a different matter: they apply only to high-risk AI systems, and only from 2 December 2027 for the Annex III categories.

Article 4: The Obligation That Is Already In Force

Article 4 has applied since 2 February 2025. The Digital Omnibus on AI (Regulation (EU) 2026/1744, in force since 27 July 2026) rewrote it. Article 4(1), as amended by Regulation (EU) 2026/1744, now reads:

"Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual."

So it is an obligation of effort: you must take measures, and you do not have to guarantee a level per person. The Act defines AI literacy in Article 3(56) and explains it in Recital 20: the skills, knowledge and understanding that let people use AI systems in an informed way and be aware of their opportunities, risks and possible harm.

Three things follow from that:

  1. Context matters. The measures for a recruiter using an AI screening tool differ from those for a customer service agent using a chatbot. The Commission's Q&A says different levels of training or learning approaches can be appropriate.

  2. It covers both opportunities and risks. Staff need to know what can go wrong, not just how to prompt effectively. The Commission gives ChatGPT users as an example: they should be informed about specific risks such as hallucination.

  3. It covers everyone using AI on your behalf. That means staff and others dealing with the operation and use of your AI systems, including contractors.

What This Looks Like in a 50-Person Company

Take a typical Dutch SME with 50 employees. The finance team uses an AI tool to flag anomalous invoices. The marketing coordinator uses ChatGPT to draft copy. The HR manager uses LinkedIn Recruiter's AI features to rank candidates.

Reasonable Article 4 measures for that company look like this:

  • Identify which staff are using AI systems as part of their role. This is your AI use inventory. Without it, you cannot scope your literacy measures.

  • Map each use case to a risk level. Using AI to sort invoices is different from using AI to rank job applicants. The latter falls under the high-risk categories in Annex III of the Act, with extra duties from 2 December 2027.

  • Design role-specific training. The HR manager needs to understand bias risks and when a human must review AI-assisted decisions. The finance analyst needs to understand false-positive rates and auditability. The marketing coordinator needs to understand output accuracy and intellectual property considerations.

  • Keep records. Article 4 does not require records, and no certificate is needed. But records are how you show what you did. The Commission suggests an internal record of trainings and other initiatives. A short log with dates, topics covered and staff names does the job.

None of this requires a legal team. It requires a documented process.

The Gap Most Companies Have Right Now

A generic "AI awareness" lunch-and-learn, or asking staff to watch a vendor webinar, is a weak basis for Article 4. The Commission's Q&A notes that simply relying on a tool's instructions for use, or asking staff to read them, is often ineffective.

The generic session ignores the context your staff work in. A vendor onboarding video is not a measure you took. And doing nothing at all has been non-compliance since 2 February 2025.

The AI Act sets no specific fine for Article 4. Article 99 leaves penalties for infringements outside its fixed tiers to national law, and Member States can choose fines, warnings or non-monetary measures. National market surveillance authorities supervise Article 4 from 2 August 2026. In the Netherlands, the draft implementing law foresees an order to fix shortcomings (a "last onder dwangsom"), not a fine, and it is not yet in force. Either way, a supervisor will want to see what measures you have taken.

Where GDPR and the AI Act Overlap

For Dutch companies, the Autoriteit Persoonsgegevens (AP) is the intended supervisor for most AI Act topics, together with sector supervisors such as the AFM and DNB for financial firms. It will only have AI Act powers once the Dutch implementing law is in force. The AP already enforces AVG/GDPR requirements around automated decision-making under Article 22 GDPR. When an AI system processes employee or customer data, you have both GDPR and AI Act obligations running simultaneously.

AI literacy training is one place where the two regimes reinforce each other. Staff who understand AI limitations are better equipped to handle data subject requests, spot unlawful automated decisions, and flag incidents. A single training programme can address both frameworks if it is designed with both in mind.

Three Steps to Take This Month

Article 4 compliance does not require a six-month project. It requires focus.

Step 1: Inventory your AI use. Ask every team lead to list the AI tools their team uses, including free consumer tools. Expect surprises.

Step 2: Classify by role and risk. Flag any use case that involves hiring, performance evaluation, creditworthiness, health, or access to essential services. These map to Annex III high-risk categories and carry heavier obligations beyond Article 4 from 2 December 2027.

Step 3: Run role-specific training and document it. Keep it short and practical. The goal is not certification; it is evidence that you took measures that fit the AI tools your people use, including what to do when something goes wrong.

These three steps position you to answer a supervisory authority's first question: "What measures have you taken to support AI literacy under Article 4?"

The answer "we sent a company-wide email" is a weak one. A training log showing role-specific sessions with documented content is a strong one.

Not sure where your company currently stands? The free 2-minute compliance check at comply.khairos.ai walks you through your key deployer obligations under the EU AI Act, including Article 4, and shows you which gaps to close first.

References

  1. EU AI Act Article 3, Definitions

  2. EU AI Act Article 4, AI Literacy

  3. EU AI Act Recital 20, AI Literacy

  4. Regulation (EU) 2026/1744, Digital Omnibus on AI

  5. European Commission, AI Literacy Questions & Answers

  6. EU AI Act Article 99, Penalties

Part of our guide to AI Literacy & Article 4

Not sure where your company stands?

Answer 15 questions about how your team uses AI. You get a compliance score and the gaps to close first, as a PDF.

Am I compliant? Check in 2 minHow Khairos Comply works