🔍 Read the full analysis: Europe’s Public Services Face A New Tool: AI on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Reports published in October describe four different approaches to AI in European government: the Netherlands lists AI among tools for reducing central-government costs, the UK uses AI in asylum casework and found errors in sampled summaries, bots accounted for most replies to a UK health consultation, and Sweden blocked public ChatGPT and Claude on government computers. The cases raise questions about staffing, decisions affecting individuals, public participation and data controls; they do not establish a single Europe-wide policy.
European governments are taking sharply different steps as artificial intelligence enters public services: the Netherlands has included AI among tools for reducing central-government costs, the UK Home Office uses AI to support asylum casework, and Sweden has blocked public versions of ChatGPT and Claude on government computers. A separate UK consultation review found that bots made up 76% of responses to a 2023 health consultation, highlighting risks to public participation as well as to government decisions.
A Dutch government progress letter sent to parliament on October 26 projects about 25,000 fewer full-time positions in central government by 2035 as spending cuts rise to €3.49 billion a year. The estimate is described in the letter as rough. AI is one of several possible measures, alongside simpler rules, shared desks and standardised IT. The government says it will set AI targets and savings plans in the first quarter of 2027, while warning against overestimating the staffing effects of digitalisation.
In Britain, the Home Office uses two tools built on OpenAI models through Microsoft Azure: one summarises asylum interview transcripts, and another searches official country information. Documents reviewed by digital rights group Open Rights Group (ORG) show that a Home Office spot check found 15% to 22% of sampled summaries inaccurate or incomplete, compared with 9% in a pilot. The department says trained caseworkers make the decisions and the tools do not decide asylum claims. Applicants are not specifically told AI was used, according to a ministerial statement cited by ORG.
ORG also says only 203 summaries were reviewed, out of an estimated 70,000 processed from April to July 2026—about 0.3%. Separately, the UK health department flagged 90,835 of 118,756 responses to a 2023 consultation as bots, according to the source report. In Sweden, the Government Offices blocked browser versions of ChatGPT and Claude on staff computers from October 26; staff are directed to approved tools. The measure concerns public services and data controls, not a ban on AI.
Infographic 01 · The Netherlands
25,000 fewer civil servants. AI is one tool, not the reason.
The driver is a budget cut. The government plans to use AI, and warns against expecting too much from it.
The cut grows every year
Cuts to the running costs of central government, € billion per year.
Police and defence are exempt. Many organisations face about 20% fewer staff, some up to 45%.
Where the money must come from
The letter’s own split of the cut.
AI is one of several tools inside both parts, next to simpler rules, standard IT, shared desks and joint purchasing.
AI in the plan
- “Ambitious targets and savings plans” for AI in Q1 2027
- Purchasing to examine “systematically” where AI fits
- Conditions: security, data quality, sovereignty
Also in the plan
- At least 500 rules a year scrapped or simplified
- One desk per two employees
- Fewer advisory bodies (28 councils, ~150 committees)
The letter’s own warning
- Past digitisation did not deliver the promised staff cuts
- It brought new tasks and more contractors
- Cuts may affect service and continuity
“Do not overestimate the staffing effects of digitalisation and AI.”Lesson 2 in the Dutch government’s own letter, 26 Oct 2026 · translation: Thorsten Meyer
Sources: BZK, Voortgangsbrief Slagvaardige Overheid + annex, 26 Oct 2026; Rijksoverheid news release, 26 Oct 2026. 25,000 ÷ 160,016 ≈ 1 in 6 and bar scaling: Thorsten Meyer.
Powered by Thorsten Meyer AI
AI Enters High-Stakes Public Work
These cases show why AI in government is more than a question of whether software can save time. It can influence the preparation of asylum case files, affect the staffing and cost of public administration, and distort the apparent level of public support for a policy. Errors or automated submissions can shape the material officials see, even when a human remains formally responsible for the decision.
The reported UK error rates matter because case summaries concern interviews in which people explain why they fled, while applicants reportedly cannot review or correct the AI-generated summary. The Home Office says caseworkers make every decision, but the low proportion of summaries checked raises questions about how errors are detected in routine use. The Dutch letter, meanwhile, cautions that AI may not produce the scale of staffing savings sometimes expected: it says earlier digitisation created new work and did not deliver promised reductions.
For residents and public employees, the immediate issue is whether governments can explain where AI is used, test its performance and provide routes to challenge mistakes. The examples are from separate national systems; they do not show that Europe has adopted one common approach.
Products Worth Considering

Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity

Amazon Kindle Scribe Colorsoft 64GB with Premium Pen

Apple MacBook Air 15-inch (2026) with M5 chip, 16GB memory, 512GB SSD — Midnight
As an affiliate, we earn on qualifying purchases.
The developments were reported together on October 29, 2026, but they arise from distinct decisions and evidence. The Dutch figure is part of a government budget and organisational plan, not a forecast that AI alone will remove 25,000 jobs. The progress letter says savings from general efficiency would cover only about a third of the planned reductions; the remainder would require decisions about which state activities to stop.
ORG published its findings on the UK asylum tools on October 12, after six months of parliamentary questions and freedom-of-information requests. The tools entered use at different points: the country-information search tool reached all decision-makers in July 2025, and the summary tool reached all caseworkers in April 2026. The government’s AI transparency register does not list either tool, according to ORG.
The consultation figures concern a different problem: automated submissions can make a public consultation appear to have more human participation than it did. Sweden’s action is different again: it restricts access to two public chatbot services on government devices, while permitting staff to use approved alternatives. The EU AI Act also matters to the wider debate: the source report says its relevant high-risk obligations for areas including asylum and access to public services are due to apply on December 2, 2027, following a legislative delay.
“Do not overestimate the staffing effects of digitalisation and AI.”
— Dutch government progress letter to parliament
Error Checks and Oversight Gaps
The reported Home Office spot check covered 203 summaries, and the source material does not establish how representative that sample was. It also does not provide a complete account of the kinds of errors found, how often they affected decisions, or whether errors were corrected before decisions were made. The department’s position is that caseworkers decide claims, but the available information does not show how consistently they identify inaccurate summaries.
The source reports that the UK’s Independent Chief Inspector of Borders and Immigration asked ORG for further information while planning next year’s inspections. That does not confirm a formal investigation. It remains unclear whether either asylum tool will be added to the public transparency register, whether applicants will be able to see or correct summaries, or whether the department will expand checks.
The Dutch 25,000-job figure is a rough estimate tied to the wider spending plan, and the government has not yet published its AI savings targets. The source also does not describe the approved AI tools available to Swedish staff or the criteria for their use. These details will affect how the policies work in practice.
Plans, Inspections and New Rules
The Dutch government says it plans to publish AI targets and savings plans in the first quarter of 2027. Those proposals should clarify what role AI is expected to play in the wider reductions and how the government will distinguish projected savings from realised ones.
In the UK, the immigration inspector’s planned work next year may bring further scrutiny of the asylum tools, although no probe has been confirmed. Any additional findings could clarify the systems’ accuracy, the frequency of human checks and whether applicants can challenge errors. The Home Office’s position remains that staff, not the tools, make decisions.
The next wider milestone is the scheduled start of relevant EU AI Act obligations on December 2, 2027, as described in the source report. Until then, national rules and departmental practices will continue to shape how these systems are used. The immediate questions for governments are whether they publish clear records, test tools against real-world use and explain how people can seek review when AI-generated material is involved.
Key Questions
Is the Netherlands planning 25,000 AI-driven job cuts?
No. The government letter estimates about 25,000 fewer full-time positions by 2035 as part of broader spending cuts. It lists AI as one of several tools and warns against overestimating its staffing effects.
Does AI decide UK asylum claims?
The Home Office says trained caseworkers make every decision and that the tools do not decide claims. One tool summarises interview transcripts and another searches official country information.
How many asylum summaries failed the UK spot check?
In a sample of 203 summaries, Home Office experts recorded a 22% failure rate in an initial review and 15% in a second review, according to ORG. The source does not establish how representative the sample was.
Why did Sweden block ChatGPT and Claude on government computers?
The Government Offices blocked the public browser versions of the services from October 26 and directed staff to approved tools. The reported measure concerns contracts and data controls; it is not a general ban on AI.
When do relevant EU AI Act obligations begin?
The source report says relevant high-risk obligations for areas including asylum and access to public services are scheduled to apply from December 2, 2027, after a legislative delay.
Source: ThorstenMeyerAI.com
