aiminute. ← All AI news
Regulation 2026-08-10

Britain's employment tribunals are drowning in claims written by chatbots — 64,000 cases waiting, some citing laws that do not exist

Britain's employment tribunals are drowning in claims written by chatbots — 64,000 cases waiting, some citing laws that do not exist

Workers in the United Kingdom who cannot afford a lawyer are drafting their own tribunal claims with ChatGPT and Grok, and the courts are buckling. The backlog of single claims reached about 64,000 at the end of March 2026, up from roughly 45,000 a year earlier, with filings up 39 per cent over the same period. The documents arriving are not short: AI-assisted claims can run to hundreds of pages and stack up dozens of separate legal grounds at once, some of them resting on statutes that do not exist. 'We get the Magna Carta, the European Convention on Human Rights and everything else thrown in,' the barrister John Bowers said. Judge Barry Clarke, president of the Employment Tribunals for England and Wales, and Judge Susan Walker, his counterpart in Scotland, issued presidential guidance that took effect on 22 June 2026 specifically to discourage a flood of interim relief applications — the emergency procedure that can order an employer to keep paying a dismissed worker — after each of the system's twelve regional offices began receiving roughly twenty a month. The Economist, which wrote the situation up on 6 August, called it a tragedy of the commons, AI edition. Labour's Employment Rights Act adds around twenty-five further grounds on which a claim can be brought and removes some compensation caps, which is expected to push the numbers higher still.

Why it mattersThis is the first large, boring, measurable case of generative AI overwhelming a public institution, and it is worth being precise about who is at fault, because the obvious answer is wrong. The people filing these claims are mostly not abusers of the system; they are workers who were sacked, cannot pay for representation, and now have a tool that will write them something that looks like a legal document. Access to justice genuinely widened. The cost is that the filter which used to sit between a grievance and a court — a solicitor telling you which of your twelve complaints is actually actionable — has been removed, and the tribunal now performs that filtering itself, at public expense, one case at a time. Expect the same pattern anywhere a form stands between a citizen and a state: benefits appeals, planning objections, immigration applications, freedom of information requests. And note the remedy the tribunals reached for. They did not ban AI-written claims, which would be unenforceable. They made one specific, heavily abused procedure harder to invoke. That is probably the shape of most workable AI regulation for the next few years: not rules about the technology, but friction added at the point where its output hits a system that cannot absorb the volume.

✓ Verified · 4 sources

WhatsApp X Telegram
Read in the app — free, in 9 languages

Related stories

Nevada just cleared up to 8,000 driverless taxis for Las Vegas — Tesla's ceiling went from 10 to 5,000 in five weeks
2026-08-21
OpenAI has started a team whose stated worry is that a company like OpenAI ends up with too much power
2026-08-21
OpenAI says it can police misuse without keeping your company's data — and it is pointing at Anthropic while saying so
2026-08-20
A Florida officer ran his ex-wife's plate 717 times. A Kansas city has just switched its cameras off.
2026-08-20
Six months ago Hollywood sent ByteDance a cease-and-desist. This week they signed a truce — and no money changed hands.
2026-08-20