AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: From 25,000 Jobs To 90,835 Replies: Europe’s Government AI Story on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

A report published on October 29, 2026, brings together three European government AI developments: the Netherlands plans a smaller central-government workforce, UK officials use AI tools in asylum casework, and a 2023 health consultation discarded most replies after they were flagged as bots. Sweden has also blocked public ChatGPT and Claude access on government computers. The cases raise questions about staffing, scrutiny of decisions affecting people, and the reliability of public consultation responses.

A report published on October 29, 2026, draws together three different questions facing European governments using AI: how much it may reduce public-sector staffing, how carefully it is checked when used in asylum casework, and how to distinguish genuine public feedback from automated submissions. The developments include a Dutch estimate of about 25,000 fewer central-government jobs by 2035, UK evidence of errors in AI-generated asylum summaries, and a health consultation in which 90,835 of 118,756 replies were flagged as bots and excluded.

The Dutch estimate appears in a government progress letter sent to parliament on October 26 by State Secretary Eric van der Burg. The letter describes a plan to reduce central-government running costs, with annual cuts reaching €3.49 billion from 2035. It calls the estimate of 25,000 full-time positions a rough calculation. AI is listed as one possible efficiency tool, alongside changes such as simpler rules and shared services; the letter cautions against assuming digitalisation will deliver large staffing reductions.

In the UK, the Home Office uses two AI tools in asylum casework: one summarizes interviews and another searches official country information. Documents obtained by the digital rights group Open Rights Group (ORG) indicate that both use OpenAI models through Microsoft’s Azure cloud. ORG reported that a Home Office check of 203 summaries found 45 failed an initial review and 30 failed a second review. The department’s pilot had recorded a 9% inaccurate-or-incomplete rate. The Home Office says trained caseworkers make decisions and the tools do not decide asylum claims.

A separate UK health consultation received 118,756 replies in 2023. The health department flagged 90,835, or 76%, as bots and discarded them, according to the source report; it says excluding those replies changed the outcome. Sweden’s Government Offices, meanwhile, blocked staff access to the public browser versions of ChatGPT and Claude from October 26, directing employees to approved tools. The report describes this as a restriction on products and data handling, not a ban on AI use.

At a glance
reportWhen: Reported October 29, 2026; developments…
The developmentA report published October 29 links three government AI developments in Europe, including a Dutch plan for about 25,000 fewer civil-service positions and UK concerns over asylum summaries and bot-flagged consultation replies.
Thorsten Meyer AI
Plan · 26 Oct 2026

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.

~25,000fewer full-time jobs in central government by 2035, a “rough estimate”: about 1 in 6
€3.49Byearly cut to running costs from 2035 (€0.90B in 2027)
+41%staff growth since 2018: 160,016 full-time equivalents at the end of 2025
Q1 2027when the cabinet publishes its AI targets and savings plans

The cut grows every year

Cuts to the running costs of central government, € billion per year.

  • 2027after €0.39B in 2025€0.90B
  • 2030€2.49B
  • 2035and every year after€3.49B

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.

  • General efficiency0.5% a year, the planning bureau’s benchmark≈ 1/3
  • Strategic choiceswhat the state does less of, or stops≈ 2/3

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

Three Tests for Public-Sector AI

Together, the cases show that government AI is not one policy choice. In the Netherlands, the central question is whether AI can contribute to savings without being credited for workforce cuts it may not produce. The government’s own letter says earlier digitalisation and automation created new tasks and contractor needs, while promised staffing savings did not materialize. It estimates general efficiency can cover only about one-third of planned reductions; the rest would require decisions about which services or activities to stop.

In asylum work, errors in a summary can matter even if a human official retains formal responsibility for the decision. Applicants are not specifically told when AI has been used, according to a ministerial statement cited by ORG, and cannot see or correct their own summary. The reported gap between pilot and later checks makes the extent of ongoing quality control relevant to people whose claims are assessed using the tools.

The consultation case raises a different accountability problem: filtering automated submissions may protect public processes from manipulation, but excluding a large share of replies can also affect what a consultation appears to show. The report says the result changed after the flagged responses were removed. Readers cannot assess that change fully without the department’s methodology and evidence about how the bot classification was validated.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

How the Cases Developed

The Dutch letter sets out cuts that rise from €901 million in 2027 to €3.49 billion annually from 2035. The figure of 25,000 is not a confirmed staffing decision for each agency: it is the letter’s rough translation of the savings target into jobs. The departments and agencies covered employed 160,016 full-time equivalents at the end of 2025, while police and defence are exempt. The cabinet says it will publish AI targets and savings plans in the first quarter of 2027.

ORG published its UK report on October 12, following six months of parliamentary questions and freedom-of-information requests. The asylum policy search tool reached all decision-makers in July 2025; the summary tool reached all caseworkers in April 2026. The report says about 17,515 cases a month pass through the summary tool, while experts checked 203 summaries in the cited spot check—about 0.3% of the estimated volume from April to July. The source material also says the EU AI Act classifies AI used in asylum decisions and access to public services as high-risk, with relevant duties due to start on December 2, 2027, following changes described as the digital omnibus.

What the Evidence Does Not Show

The Dutch figure is an estimate tied to a budget plan, not a final count of jobs that will be removed. The letter does not establish how much of the reduction, if any, will result directly from AI. The scope of future service changes and the agencies’ eventual staffing plans remain to be set out.

For the UK asylum tools, the reported sample does not establish the error rate across every summary or show how often errors affected a final decision. The Home Office’s sampling and correction procedures, and whether applicants will be given notice or a chance to challenge summaries, are not clear in the supplied material. The source says the tools are not listed in the government’s algorithmic transparency register, but does not provide a Home Office response on that omission.

For the health consultation, the supplied account does not explain the department’s criteria for identifying bots, whether the flagged replies were independently audited, or the precise nature of the changed outcome. Sweden’s restriction is confirmed as applying to public versions of two services on staff computers; details of the approved alternatives and their safeguards are not provided.

Upcoming Reviews and Government Plans

The Dutch cabinet is expected to publish its AI targets and savings plans in the first quarter of 2027, offering more detail on how AI fits into the wider spending reductions. Those plans should clarify whether the job estimate is revised and what services or tasks may be affected.

In Britain, WIRED reported on October 27 that the Independent Chief Inspector of Borders and Immigration had asked ORG for further information while planning inspections for next year. The inspector had not confirmed an investigation, so whether the tools will receive a dedicated review remains unresolved. The Home Office’s use of the tools and any further accuracy checks will also be relevant to assessing ORG’s findings.

The source material provides no timetable for a review of the health consultation’s bot classifications or for changes to Sweden’s approved-tool policy. More detail on verification methods, disclosure to affected people and public reporting would help establish how each government is testing AI systems that influence staffing, individual cases or public input.

Key Questions

Is the Netherlands cutting 25,000 jobs because of AI?

Not specifically. The Dutch government letter estimates about 25,000 fewer full-time positions by 2035 as a rough translation of broader cost reductions. It lists AI among several possible efficiency tools and warns against overestimating AI’s staffing effects.

Does AI decide UK asylum claims?

The Home Office says trained caseworkers make every decision and that its tools do not decide claims. One tool summarizes interviews and another searches country information; ORG has raised concerns about accuracy, disclosure and the limited number of summaries checked.

How many health consultation replies were flagged as bots?

The health department flagged 90,835 of 118,756 replies to a 2023 consultation—76%—as bots and excluded them, according to the source report. The supplied material does not detail how the classification was validated.

Has Sweden banned AI in government?

No. From October 26, Sweden’s Government Offices blocked staff from using the public browser versions of ChatGPT and Claude on work computers. Staff were directed to approved tools; the source describes a restriction on products and data handling, not a general AI ban.

Source: ThorstenMeyerAI.com

You May Also Like

A Closer Look At Claude’s 35,000 Designs For Allergy Drug Research

Anthropic says Claude generated 35,000 designs in allergy drug research, but the available report does not say what they were or whether they were tested.

A Look At 9 AI Mini PCs For Local Model Workloads In 2027

A 2027-oriented roundup compares nine mini PCs by memory, storage, processors and ports, while noting that performance claims lack benchmark evidence.

Claude On Google Cloud: Availability Meets Deployment Flexibility

Google Cloud says more organizations can access Claude, but regional support, pricing and deployment details have not been specified.

OpenAI’s In-Software Agent Training Brings Ironclad’s Terms Into View

OpenAI says GPT-6 Astra met 55% of rubric criteria across 11 Ironclad tasks; time figures are simulated, and human review remains necessary.