Felicity Harber v The Commissioners for HMRC
Decision date: 4 December 2023
Neutral citation: [2023] UKFTT 1007 (TC)
Overall AI summary confidence: high
Short overview
This short overview is intended to summarise the case, issues and outcome so far as they are supported by the judgment.
AI confidence in this short overview: high
This appeal concerned a 20% failure-to-notify CGT penalty of £3,265.11 after Mrs Harber sold a property and did not notify HMRC by the statutory deadline. The Tribunal found nine FTT decisions relied on by the appellant were fabricated AI-generated summaries, rejected her mental health and ignorance-of-law defences after applying the Perrin test, and dismissed the appeal, confirming the penalty. The decision records that parties should verify authorities and may not evade penalties by relying on inauthentic or unverified case law.
Ratio decidendi
This summary is intended to identify the ratio decidendi, meaning the legal reasons for deciding and the binding part of the decision.
AI confidence in this ratio decidendi summary: high
The Tribunal applied the Perrin four-stage approach and held that a taxpayer’s personal circumstances, including mental health, must be assessed objectively against their demonstrated conduct; where the taxpayer nonetheless took significant related steps (contacting HMRC, engaging accountants, setting aside funds) the circumstances did not amount to a reasonable excuse. The Tribunal also treated reliance on authorities that are not authentic (including AI-generated fabrications) as unacceptable and a proper ground for criticism, requiring parties to verify the authenticity of cases they rely on.
Obiter dicta
This summary is intended to identify obiter dicta, meaning observations made by the way that were not necessary to deciding the case and are not binding.
AI confidence in this obiter dicta summary: high
The judgment comments that submission of fabricated judgments wastes tribunal and party resources and harms the administration of justice, and warns that AI language models can "hallucinate" plausible but incorrect case summaries, so parties should be cautious and verify AI-produced authorities. These observations are expressed as broader remarks rather than necessary grounds for the outcome.