AI literacy and certification in government.
Government is the sector where AI is most visibly under political and legal pressure. In early 2026, approximately 1,392 algorithms were listed in the Algoritmeregister.overheid.nl, while research by the Algemene Rekenkamer covering more than 70 institutions showed that only 5% of actively used AI systems had been registered. AI literacy here is not a technical matter but a matter of administrative law.
What is at play
What is at play in the public sector.
At UWV, Politie, municipalities and executive agencies, AI applications have been widely introduced. UWV has nine registered algorithms, including the Risicoscan Verwijtbare Werkloosheid. Belastingdienst and Dienst Toeslagen use risk models that, according to the Rekenkamer (May 2025), do not comply with the GDPR. Amsterdam leads among municipalities with approximately fifty registered algorithms. Rechtspraak launched a national AI programme in June 2025. The widely reported case of the cantonal judge in Gelderland who used ChatGPT in a judgment (6 August 2024) demonstrated that these risks are real and immediate.
Why AI literacy matters especially here
Profiling, fundamental rights and high-risk.
The toeslagenaffaire and the SyRI ruling of 5 February 2020 exposed the sensitivity of profiling and risk classification. The AI Act (Annex III) classifies AI for law enforcement, migration, access to essential public services and biometric identification as high-risk. In addition, the GDPR, the Algemene wet bestuursrecht, the Wet open overheid and the Archiefwet apply. The updated Impact Assessment Mensenrechten en Algoritmen (IAMA, February 2026) aligns with the FRIA requirement under Article 27 of the AI Act.
The four SAIG levels for the public sector
Four levels, one framework.
For all employees who occasionally interact with AI, for example via public-service chatbots.
For frontline civil servants who regularly use AI: case handlers, permit officers, inspectors in first-line roles.
For policy officers, lawyers, enforcement officials and data analysts who must independently assess AI output.
For algorithm coordinators, CIOs, lawyers, DPOs, municipal secretaries and executives responsible for IAMA/FRIA, registration and governance.
Which level fits which role?
Role and recommended level.
| Role | AI contact | Recommended level |
|---|---|---|
| Receptionist KCC, counter employee | chatbots, general information | Awareness Badge |
| Benefits case handler, permit officer | daily AI in casework | Basis |
| Policy officer, lawyer, inspector | independently assesses AI outcomes | Practitioner |
| Data analyst, enforcement officer with risk model | works directly with model output | Practitioner |
| Algorithm coordinator, CIO, CISO | governance, registration, IAMA | Advanced |
| Executive, alderman, municipal secretary, DPO | political and legal final responsibility | Advanced |
For organisations and for professionals
One standard, two tracks.
For public-sector organisations, SAIG provides a uniform, independent competence framework that aligns with the Algoritmekader (BZK), IAMA and the intended supervision by AP and RDI. This links demonstrable AI literacy (Article 4) to human oversight of high-risk AI (Article 14, Article 26(7)). For individual civil servants, the certificate is portable across ministries, municipalities and executive agencies — verifiable proof of tested competence, not of participation.
Frequently asked questions
Frequently asked questions about AI in the public sector.
Is a SAIG certificate mandatory for civil servants?
No. Article 4 requires a sufficient level of AI literacy; SAIG certification is one way to demonstrate this.
How does SAIG relate to IAMA and FRIA?
IAMA and FRIA are assessments at the system level. SAIG applies to the individual. Anyone conducting an IAMA or FRIA should be literate at least at Practitioner level; those with final responsibility at Advanced level.
Must an alderman or executive board member be certified?
Not by law. SAIG-Advanced is however appropriate for executives responsible for procurement, policy and public accountability of AI.
Which level fits a policy officer who uses ChatGPT?
Typically Practitioner: output must be independently verified before it ends up in documents.
Does SAIG replace registration in the Algoritmeregister?
No. The Algoritmeregister documents systems; SAIG certifies individuals.
Who supervises compliance?
From 2 August 2026, AP and RDI are the coordinating market supervisors, alongside sectoral inspectorates.
Next step
Schedule an orientation call.
We discuss what Article 4 means specifically for your organisation in the public sector and which SAIG route fits your roles and risk profile.
Sources: Algoritmeregister.overheid.nl; Algemene Rekenkamer (21 May 2025); AP and RDI final advice coordinating supervision; IAMA (February 2026); Rechtspraak AI strategy; EUR-Lex Regulation (EU) 2024/1689.