Taxpayer High-Income Child Benefit AI Appeal Failed

The use of artificial intelligence in legal research and tax dispute preparation is increasingly prevalent, offering speed, convenience, and cost savings. However, a recent UK First-tier Tribunal case highlights the dangers of relying solely on AI without human oversight. This case involved a taxpayer contesting a High-Income Child Benefit Charge (HICBC) assessment, demonstrating important lessons about self-assessment obligations, AI limitations in litigation, and the legal grounds of HMRC’s discovery powers.

The Background

For the 2018/19 tax year, the taxpayer—let’s call him Mr Z—had an adjusted net income of £ 0,000 and was receiving child benefit. That combination automatically tripled liability for the HICBC, a tax charge designed to recover child benefits from higher-earning households.

However, Mr Z did not file a self-assessment return for that year, nor did he notify HMRC of his HICBC liability. Under self-assessment rules, individuals who owe the charge must report it—even if all other income is taxed through PAYE.

In April 2021, HMRC identified the omission and calculated the unpaid charge. In January 2023, they issued a discovery assessment to recover the tax due. Mr Z appealed to the FTT, claiming he should not have to pay the charge.

The AI Factor

Representing himself, Mr Z admitted using AI to research relevant case law and prepare his statement of case. He explained that he lacked the technical skills to conduct detailed legal research, so AI seemed like an efficient solution.

The tribunal accepted that using AI was a logical choice for a self-represented appellant, especially as it can help identify potential arguments and past decisions. However, they stressed a crucial point—AI is a tool, not an authority. All output must be verified for accuracy, relevance, and applicability.

In Mr Z’s case, while some legal citations were genuine, others were misquoted or irrelevant to his circumstances. The tribunal noted Z’s submissions contained propositions unsupported by the cases he cited.

The Arguments

Mr Z’s case put forward several main lines of defence:

1. The Wilkes Precedent: He argued that earlier legal decisions on HICBC should have prevented HMRC from issuing a retrospective charge for 2018/19.
2. Fairness and Complexity: He claimed that calculating adjusted net income is too complex for employees on PAYE with variable income, and that HMRC should proactively inform taxpayers of their obligations.
3. Human Rights Concerns: He suggested the HICBC was unfair and possibly in breach of his rights.
4. Legislative Challenge: He criticised Section 97 of the Finance Act 2022, which clarified HMRC’s power to backdate HICBC charges, calling it unjust.

Why the Tribunal Rejected the Appeal

The FTT dismissed all of Mr Z’s HMRC’s claims. Their reasoning highlights several important legal points:

1. Wilkes Did Not Apply: The tribunal found that the cited precedent did not cover Mr Z’s circumstances, partly because of timing—he had not appealed within the relevant legal windows.
2. Complexity Is No Defence: The self-assessment regime places the onus on taxpayers to understand and report their liabilities. The tribunal confirmed that HMRC is not obliged to issue personalised reminders or notifications of HICBC liability.
3. Legislation Is Binding: Even if Mr Z believed Section 97 of the Finance Act 2022 to be unfair, the tribunal’s role is to apply the law, not rewrite it.
4. Discovery Assessment Was Valid: Under the Taxes Management Act, a discovery assessment is valid if:
   • The officer subjectively believes there is a loss of tax.
   • That belief is objectively reasonable based on the facts.
   • The tribunal found both conditions satisfied. Mr Z’s undisputed income level and receipt of child benefit were sufficient grounds for HMRC to act.

Lessons on AI in Legal Disputes

This case does not criticise AI in tax litigation. The tribunal acknowledged that AI can be useful, especially for people who don’t have professional help. However, it highlights an important limitation: AI-generated content needs to be checked for factual accuracy and legal relevance.

AI can misunderstand the details of a case, make wrong comparisons, or refer to outdated laws. If these mistakes are not carefully reviewed, users might mistakenly believe their situation is stronger than it really is.

Broader Takeaways for Taxpayers

1. Self-Assessment Is Self-Responsibility: Even if you are a PAYE employee, certain situations—such as earning above £50,000 while receiving child benefit—trigger a legal duty to report and pay extra tax.
2. HMRC’s Discovery Powers Are Strong: Once HMRC discovers an underpayment, they can issue assessments that are enforceable as long as the legal tests are met. Challenging these requires clear evidence of procedural error or legal misapplication—not general objections to fairness.
3. AI is a Starting Point, Not an Endpoint: Whether you are an individual or a professional adviser, AI tools should be used as research assistants, not as sole authorities. Always cross-check with the legislation and authoritative case reports.
4. Deadlines Matter: Even strong arguments can be lost if procedural time limits are missed. Appeals must be lodged promptly, and challenges to HMRC assessments must be raised within the statutory windows.

Best Practices for Using AI in Tax Cases

For those considering AI in case preparation:

• Verify Citations: Use official legal databases to confirm the details of any case law AI suggests.
• Context-Check: Ensure the cases actually apply to your situation.
• Stay Current: AI can pull outdated legislation or pre-amendment rules—check the latest versions.
• Pair with Human Expertise: Use AI to draft outlines, but seek a professional’s review before submission.
• Don’t Over-Rely: Keep AI as a supplement to, not a substitute for, sound legal reasoning.

Mr Z’s experience is a cautionary tale. AI may have streamlined his research, but without critical evaluation, it led him to rely on arguments the tribunal could not accept. The discovery assessment stood, and the HICBC liability remained.

The case reinforces two enduring truths in UK tax law:

• Taxpayers must take active responsibility for meeting their obligations, regardless of complexity.
• Technology can assist, but cannot replace, the careful application of law to fact.

In an age where AI promises to “democratise” legal research, this decision is a reminder that winning a tax appeal still depends on accuracy, relevance, and a “firm grasp” of the rules.