Britain’s Bureaucratic Paper Trail Is Growing With AI

🔍 Read the full analysis: Britain’s Bureaucratic Paper Trail Is Growing With AI on ThorstenMeyerAI.com

Written complaints and claims to British public bodies have risen across multiple services since ChatGPT launched in November 2022, according to a Sky News report published October 20. The report links the pattern to the lower effort of using AI to draft formal submissions, while stressing that the figures do not prove AI caused the increases and that many claims appear to reflect genuine problems.

Sky News technology correspondent Rowland Manthorpe reported growth across a range of channels, including complaints about water companies, local authorities, GP practices and the Department for Work and Pensions, as well as individual employment tribunal claims. The outlet’s figures show household complaints about water companies up 144% since ChatGPT’s launch, complaints to English local authorities up 78%, and written complaints to GP practices up 71%. It also reported increases in DWP complaints of 64% and individual employment tribunal claims of 62%.

Other channels also rose, though by smaller reported amounts: legal-services complaints were up 49%, written complaints to the NHS 35%, police complaints 34%, and complaints about public services 23%. These are growth figures reported by Sky News; the source material does not provide the underlying totals or enough detail to establish a single comparable measurement period for every channel. The report says each service has its own possible causes, including specific pressures on water companies, and treats the repeated pattern—not any one figure—as the notable finding.

Manthorpe also tested 100 recent FOI requests with an AI-text detector. It labelled 89% as fully AI-written and 9% as human-written; a smaller sample of requests from 2019 was labelled entirely human. Sky News cautioned that AI detectors are unreliable and that the test does not establish why FOI requests increased. Separately, a survey of civil servants cited by the report found that 58% said AI-generated public submissions significantly affect their work.

More Claims Meet Paper-Based Services

The rise matters because many public services still require people to explain a problem, make a case or appeal through written forms and letters. AI tools can make that process easier for people who struggle to write formal submissions, but a larger volume also means officials must handle more material. Civil servants surveyed described receiving correspondence at speed and said some exchanges felt like “arguing with ChatGPT,” reflecting reported pressure on staff rather than proof that submissions are invalid.

The central policy choice is whether to add barriers to submissions or change how services process them. Sky News reported that the share of DWP complaints upheld has fallen from the 50% range to close to 40%. That is a lower proportion, but it does not mean fewer justified complaints in absolute terms when the total number has grown. The report’s figures point to a practical risk: rules aimed at discouraging AI-assisted submissions could also make it harder for people with valid complaints to be heard.

Former Government Digital Service founder Tom Loosemore argues that the deeper issue is reliance on essays and written explanations. He said the trend is “pretty existential” for services that depend on applicants writing extensive material. Reworking processes around structured information could reduce avoidable paperwork while preserving routes to challenge decisions. The report does not set out a tested replacement model or quantify the costs of such changes.

The Rise Since ChatGPT’s Launch

Sky News described a long period in which FOI request numbers sent to British public bodies changed little, followed by a gradual rise after ChatGPT became publicly available in November 2022 and faster growth later. The report says a similar broad shape appears across unrelated complaint and claims systems. That timing is consistent with wider use of generative AI, but timing alone cannot show that AI caused the rise.

The UK Information Commissioner’s Office issued guidance in May after public authorities reported higher volumes and more complex AI-generated FOI requests, some of which misstate the law or require clarification. The ICO’s guidance, as summarized by Sky News, says an AI-written request remains valid; authorities do not need to determine whether a person or AI wrote it, and they can seek clarification when a request is unclear.

The report also places the UK figures alongside international examples described in research by Chris Schmitz and reported by TechCrunch in September. Schmitz identified 84 possible cases of “agentic flooding” across 11 jurisdictions. The source says the UK Housing Ombudsman received about 2,600 complaints in 2022 and just over 7,000 last year, while complaints to the US Consumer Financial Protection Bureau grew fivefold. These comparisons are reported examples, not proof that all of the growth arose from AI.

““It’s like being bombarded with correspondence at speed.””

— Civil servants in an academic survey cited by Sky News

What the Figures Cannot Prove

The reported increases do not isolate AI as the cause. Changes in service performance, public awareness, access to advice, rules, or other circumstances could contribute, and the source material does not provide a full causal analysis or complete underlying datasets. The FOI detector exercise was small and used a tool Sky News described as unreliable, so it cannot establish how many requests were written with AI.

It is also unclear how much of the added work consists of valid claims, duplicate or poorly formed submissions, or cases that require substantial staff time. The DWP upheld-share figure is reported as having declined to close to 40%, but the source does not provide the exact counts or period needed to calculate the number of upheld complaints. The effect of AI agents that submit claims automatically remains prospective: the report demonstrates an agent service but does not document widespread automated filing at scale.

Services Face the Next Filing Wave

The next development to watch is whether public bodies adapt their processes as AI tools move from drafting messages to taking actions for users. Manthorpe raised examples such as automated parking-ticket appeals and Delay Repay rail claims. The report presents these as scenarios, not established patterns of mass automated claims.

For now, the ICO’s guidance sets out how authorities should handle AI-written FOI requests: treat them as valid, and ask for clarification when needed. The report leaves open whether governments and service providers will redesign applications around structured data, change staffing and processing systems, or introduce additional restrictions. Any policy response will have to account for both administrative workload and the possibility that easier submission helps people pursue claims to which they are entitled.

Key Questions

What did the Sky News report find?

It reported that complaints, FOI requests and employment tribunal claims have risen across several UK channels since ChatGPT launched in November 2022. Growth varied by service, and the report does not establish AI as the cause.

Does the report prove AI caused the increase?

No. The timing and pattern are consistent with AI making submissions easier, but other factors may have contributed. Sky News also said its AI-detector test was not conclusive.

Are AI-written FOI requests valid?

According to ICO guidance summarized by Sky News, an AI-written request remains valid. Authorities do not have to identify whether AI was used, but they can ask for clarification if a request is unclear.

Are the extra complaints mostly genuine?

The report cites researcher Chris Schmitz’s view that most cases in his research involved people claiming something they were entitled to claim. That finding does not establish the status of every complaint in the UK figures.

What changes could public services make?

Tom Loosemore has argued for reducing reliance on long written submissions and using more structured data. The report does not identify a government decision or a specific system redesign that has been adopted.

Source: ThorstenMeyerAI.com