🔍 Read the full analysis: Is AI Changing Government Work In Europe? Jobs, Asylum And Bot Replies 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
Developments reported across Europe show AI being considered for government efficiency while raising questions about staffing, asylum casework, public consultations and data controls. The Netherlands estimates about 25,000 fewer central-government jobs by 2035 but cautions against overestimating AI’s role; UK watchdog scrutiny of asylum summaries may follow, while Sweden has blocked public ChatGPT and Claude on staff computers.
Three recent government developments show how AI is entering public administration across Europe: the Netherlands estimates central-government staffing could fall by about 25,000 full-time positions by 2035, a UK rights group has questioned error rates in AI-generated asylum interview summaries, and Sweden has blocked public versions of ChatGPT and Claude on government staff computers. The cases differ, but raise immediate questions about public-sector jobs, decisions affecting applicants and the reliability of public input.
The Dutch estimate appears in a progress letter sent to parliament on October 26 by State Secretary Eric van der Burg. It translates planned operating-cost cuts into a “rough estimate” of around 25,000 full-time positions by 2035. The letter lists AI as one possible efficiency tool alongside simpler rules, shared desks, standardised information technology and joint purchasing. It also warns the government not to overestimate the staffing effects of digitalisation and AI, citing earlier efforts that did not produce promised savings.
In Britain, the Home Office uses two generative AI tools in asylum casework, according to documents obtained by digital rights group Open Rights Group (ORG). One summarises interview transcripts; the other searches official country information. ORG says the summary tool’s error rate was higher in a Home Office spot check than in its pilot: 15% to 22% of sampled summaries failed review, compared with 9% in the pilot. The department said trained caseworkers make decisions and the tools do not decide claims. Applicants are not specifically told AI was used, according to a ministerial statement cited by ORG.
A separate analysis of a 2023 UK health consultation found that the Department of Health and Social Care flagged 90,835 of 118,756 replies—76%—as bot submissions and excluded them. The source report says the consultation’s result changed when those replies were removed. Sweden’s Government Offices, meanwhile, blocked browser versions of ChatGPT and Claude from October 26, directing staff to approved tools. The reported measure concerns use of those public services on work computers, not a general 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’s Role in Public Services
Together, the developments show that public agencies are managing more than a technology rollout. They must weigh staffing and service capacity against budget targets, and protect people when AI tools support work involving asylum claims or access to services. A summary that omits or misstates details could matter even if a human official retains formal responsibility for the decision.
The consultation findings raise a different governance problem: public agencies may receive large volumes of automated submissions, but excluding them can affect how a response is interpreted. The available figures describe one consultation and do not establish how common bot submissions are across government consultations.
Sweden’s restrictions underline that adoption is also a question of data handling and procurement. Governments may permit approved systems while blocking consumer versions whose terms or data protections do not meet workplace requirements. These are distinct policy choices, not evidence that governments have settled on a single approach to AI.
Products Worth Considering
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
Three Different Government Uses
The Dutch plan is first a budget-reduction programme, not a stated target to replace 25,000 workers with AI. The letter says annual operating-cost reductions are scheduled to rise from €901 million in 2027 to €3.49 billion from 2035. It estimates that general efficiency may cover about one-third of the cuts; the rest would require decisions about which government activities to stop. Police and defence are exempt from the reductions described.
In the UK asylum system, the tools have separate functions. The Asylum Case Summarisation tool condenses interview transcripts, while the Asylum Policy Search tool provides chat-style searches of official country information. ORG says the tools use OpenAI models through Microsoft Azure. The Home Office says the systems support caseworkers rather than make asylum decisions. ORG also reports that only 203 summaries were checked against an estimated roughly 70,000 summaries processed from April through July—about 0.3% of that estimated volume.
The EU AI Act classifies certain AI uses connected to asylum decisions and access to public services as high-risk. The source report says that, following this year’s “digital omnibus,” the relevant duties are due to begin on December 2, 2027, rather than in 2026. The three national examples therefore sit within a wider European debate about oversight, transparency and when public bodies should use AI.
“Do not overestimate the staffing effects of digitalisation and AI.”
— Dutch government progress letter to parliament, sent by State Secretary Eric van der Burg
Oversight and Error Rates
The reported UK spot check covered 203 summaries, and the source material does not provide details about how those cases were selected or how the two review rates were calculated beyond the stated first and second reviews. It is not clear whether the Home Office will expand checks, publish further accuracy data or change the tools. ORG says applicants cannot see or correct the summaries, but the source material does not describe a planned process for giving them access.
On October 27, WIRED reported that the Independent Chief Inspector of Borders and Immigration had asked ORG for more information while planning next year’s inspections. The inspector had not confirmed an investigation. The source material also refers to a July upper tribunal judge’s statement but cuts off before giving its substance, so no further conclusion can be drawn from it here.
The Dutch figure is an estimate of jobs associated with a wider programme of cuts, not a confirmed count of positions to be eliminated specifically through AI. The government’s AI savings targets and plans are due in the first quarter of 2027. For the health consultation, the available account does not identify how bot replies were detected or give the consultation’s final policy outcome.
Plans, Inspections and Rules
The Netherlands says it will publish AI targets and savings plans in the first quarter of 2027. Those documents should clarify which government functions might use AI and how the cabinet expects technology to contribute to the wider cost reductions.
In Britain, next year’s inspections by the Independent Chief Inspector of Borders and Immigration may provide a further opportunity to examine the asylum tools, but a probe has not been confirmed. The Home Office’s checking practices, transparency about AI use and the treatment of errors remain points to watch. Across the EU, the reported high-risk AI duties are scheduled to take effect on December 2, 2027.
Sweden’s staff restrictions are already in force for public ChatGPT and Claude browser versions, according to the report. The source material does not give a timetable for a review or say whether the Government Offices will change its list of approved services.
Key Questions
Is the Netherlands planning to replace 25,000 civil servants with AI?
No. The government letter gives a rough estimate of about 25,000 fewer full-time positions linked to a wider programme of operating-cost cuts by 2035. It lists AI as one possible efficiency tool, but also warns against overestimating its staffing effects.
Does AI decide UK asylum claims?
The Home Office says trained caseworkers make every decision and the tools do not decide claims. The tools described support casework by summarising interview transcripts and searching official country information.
How many UK asylum summaries failed the reported check?
In a Home Office spot check of 203 summaries, 45 failed the first review and 30 failed the second, according to figures cited by Open Rights Group. The source report gives rates of 22% and 15%, respectively, compared with a 9% failure rate in the pilot.
Did Sweden ban AI in government?
No general AI ban is described. The Government Offices blocked the browser versions of ChatGPT and Claude on staff computers from October 26, while staff were directed to approved tools.
When do the relevant EU AI Act duties start?
The source report says the duties for the specified high-risk uses are scheduled to begin on December 2, 2027, following this year’s digital omnibus.
Source: ThorstenMeyerAI.com
