Using AI in tax appeals can be problematic if it replaces legal judgment. Representing yourself at the First Tier Tribunal is one thing, but relying entirely on AI to challenge an HMRC decision, without human oversight, is another matter entirely.
Discovery Assessment
A recent case involved a discovery assessment for just over £2,500, related to the high-income child benefit charge in 2018/19. Delays in the case were due to ongoing questions about HMRC’s ability to raise discovery assessments. However, the law has now been clarified in HMRC’s favour.
Legislation was passed confirming that these assessments are valid, even retrospectively. So, arguments based on procedural grounds will no longer hold up in such appeals.
Artificial, Not Entirely Intelligent
The taxpayer’s defence was weakened by excessive dependence on AI. For instance:
- They claimed HMRC should have informed them of the charge, even though the responsibility for reporting lies with the taxpayer.
- AI-generated case law was cited, but the cases, while not fabricated, were irrelevant to the actual issue.
This highlights the main issue with using AI in tax appeals: AI may not fully understand the context or pick up on what truly matters. It can produce answers that sound convincing but miss the mark entirely.
The judge acknowledged that using AI for legal research is acceptable—but only if the output is carefully reviewed by a human.
HMRC’s Use of AI
HMRC uses AI extensively, particularly for analysing data patterns. One key area is detecting mismatches between declared income and lifestyle indicators.
For landlords, this is especially important. HMRC’s AI systems can cross-reference Land Registry records, letting platforms, and tenancy deposit schemes to identify unreported rental income.
While the government’s AI playbook targets public sector use, it offers a valuable reminder: AI tools have limits and must be used responsibly.