AI is driving up complaints to public authorities: overhaul processes or face collapse

Making a complaint is easier, although the guidance can be wrong

Preparing a complaint requires identifying the competent body, explaining what happened and setting out a request. AI helps overcome those difficulties. More accessible channels also help: Datatilsynet, the Norwegian data protection authority, says its digital form also makes it easier to submit complaints.[1]

The content can improve too. According to Digi, Hege Farnes Hildrum, head of NAV's complaints body, finds that AI helps people present their arguments more clearly, although it can also make submissions unnecessarily long.[2] In Juristen, tax lawyer Liv Øvstedal Mykkeltveit highlights more precise and carefully prepared enquiries.[3]

Guidance does, however, have consequences when it goes wrong. The European Ombudsman believes assistants are helping to bring more people to the institution, including with matters outside its remit.[4] The European Data Protection Supervisor sees an apparent link between tools that misdirect users and a rise in inadmissible complaints. Its response includes redesigning the complaints website.[5]

Finding an office to approach is not enough. If it is the wrong one, the office must examine the submission and clarify its remit, while the citizen still has not reached the body that can resolve the problem.

Offices are already recording the increase

Complaints from individuals received by Datatilsynet rose from 902 to 1,848 between the two reporting years compared in its annual report.[1]

In May of the first year, it identified four complaints it considered to have been produced wholly or partly with AI. In May of the following year, there were 54. These are indications identified by the authority, with no published detection method.[1]

Sivilombudet, which oversees public administration in Norway, received 5,893 complaints and enquiries in 2025. Its report mentions administrative delays and the possible effect of AI among the explanations for the increase.[6]

In the judicial sphere, a separate category from administrative complaints, single claims before employment tribunals in Great Britain increased by 39% in the 2025/26 financial year.[7]

Guidance from the tribunals' presidents describes a particularly marked change in applications for urgent interim measures, known as interim relief. It estimates that perhaps twenty applications a year used to be received across Great Britain. Now, in their assessment, most offices receive similar numbers each month.[8]

In Scotland, appeals to the information commissioner rose by 83% in the 2025/26 financial year.[9] This figure concerns appeals, rather than all requests for access to information.

In Spain, complaints received by the Spanish Ombudsman increased by 10.9% in 2025, according to a calculation based on its annual statistics.[10] Ángel Gabilondo cited AI assistance with drafting among the causes, according to Servimedia.[11]

The European Ombudsman handled 54% more complaints that same year.[4] These are different bodies and procedures, but all show increased activity to deal with.

The authorities attribute part of the increase to AI, and no one has measured how much is due to AI alone. The increase is real. The available data do not yet show an exponential trend.

Wider use of AI points to more written submissions

If these tools become more widespread, written complaints can be expected to keep increasing. More people will be able to prepare their arguments and request a review with less difficulty. That expectation already informs some bodies' planning.

In evidence to the UK Parliament, representatives of the Parliamentary and Health Service Ombudsman, the PHSO, anticipated annual growth of more than 25% in cases requiring investigation.[12] This was the institution's forecast.

The issue thus shifts from the volume received to the ability to respond. Faster drafting does not, by itself, reduce the time needed to check facts, read documents and give reasons for a decision.

Without more capacity, the workload becomes a queue

Faced with this pressure, public authorities have three options: let the workload exceed their capacity, recruit staff or also use AI to handle cases more efficiently. Documented responses illustrate these tensions.

The PHSO described receiving twenty-page challenges half an hour after communicating a decision.[12] The speed of the new submission does not remove the duty to examine it. If it requires another review, it takes up more staff time.

In employment tribunals, the judicial guidance warns that the documentation and urgency of these hearings displace other hearings.[8] Pressure on one procedure thus affects people who were already waiting.

Scotland announced a measure to limit the number of a person's appeals investigated simultaneously to five, on a case-by-case basis where harmful patterns arise. The remaining appeals wait, and priorities are agreed with the applicant. At the time of the announcement, the measure had not yet been applied.[13]

This approach allocates limited capacity, but leaves cases waiting. The risk of collapse arises when the system accumulates more work than it can resolve and delays generate further complaints. Sivilombudet also identifies delays as a factor.[6]

Recruitment and AI can increase the capacity to respond

Recruitment offers another response. The PHSO told Parliament that it continued to review challenges and was strengthening its legal team.[12] Adding human capacity helps deal with work that does not disappear simply because it was prepared with an automated tool.

The third option is to apply AI within public administration. In England, the LGSCO, the ombudsman that handles complaints about local government and social care, launched an assistant that explains how to complain and which matters it can deal with. The institution keeps decisions in human hands.[14]

In Norway, UNE, the immigration appeals body, announced a pilot to help draft background sections. It excludes legal assessment and retains review of each decision by at least two people.[15]

The European Ombudsman reports using AI for supporting tasks such as summarising and working with documents. Its policy rules out AI deciding the admissibility, priority or outcome of complaints.[16]

These interventions address specific tasks: providing guidance, preparing background information and handling documents. Their usefulness must be measured by the work they save and the corrections they require.

Faster case handling requires safeguards to be maintained

Within its scope, GDPR Art. 22 regulates solely automated decisions that have legal or similarly significant effects, subject to exceptions and safeguards. An AI-assisted summary does not automatically fall into that category. The European institutions have their own data protection regime.[17]

In Spain, Ley 39/2015 maintains the duty to issue and notify decisions within its scope.[18] Receiving more AI-assisted submissions does not in itself create an extension. Time limits and extensions depend on the applicable procedure.

Overhauling processes is essential to avoid collapse

The combination of more incoming submissions and limited capacity to respond makes a thorough overhaul of administrative processes essential. That overhaul starts with accessible forms that separate the facts, the remedy sought and the documents. It continues with guidance on the competent body and human assistance for those who need it.

It also requires measuring what comes in: new cases, repeat submissions, misdirected submissions and response times. This information helps determine where staff are needed, where guidance is failing and which tasks can be simplified.

Where AI is involved, there must be an identifiable person responsible and human review capable of correcting the result. Citizens need to be able to point out an omitted document or a misunderstood request.

Making it easier to complain means preparing to respond. If the capacity to complain increases and public authorities do not overhaul their processes so they can resolve cases, that gap becomes a queue. If the queue keeps growing, the result is collapse.