AI in the Dock
Trade mark Claimants have received a clear message from the late summer and early fall: while AI can assist with drafting, if you rely on it irresponsibly, you run the danger of losing credibility and, in the worst situations, your appeal. The Appointed Person issued a harsh warning about the risks of depending on unverified AI-generated content after a recent UKIPO appeal exposed misuse of AI in a party’s submissions. As part of a wider discussion in the UK over the judicial use of AI, brand owners and advisors should now anticipate tribunal scrutiny of the preparation and supporting documentation of submissions.
The headline: AI-assisted writing fouled up an appeal
Practitioners noted in September a UKIPO trade mark appeal that was unsuccessful due to issues attributed to AI-assisted writing. The Appointed Person warned litigants who allowed tools outstrip their judgement and pointed out instances of “mis-use” of AI in argument preparation, which were evident in the skeleton argument and supporting documentation. The signal is obvious even though the judgment depends on the specifics: if AI contributes false information, overconfident claims, or mis citations that you are unaware of, it is your argument that wins.
The following are immediate and useful lessons for brand owners:
• Provenance is important. Tribunals want to know that you have read the sources of the facts and authorities. If not validated, artificial intelligence summaries or fabricated citations can be lethal.
Appeals are won by evidence, not by rhetoric. A weak record won’t be saved by a beautifully worded AI-assisted submission that ignores evidentiary gaps (real usage, confusion, reputation, and bad faith).
• Professional responsibility endures. The human signatory is expected to support the correctness of filings by courts and tribunals. The claim that “the tool accomplished it” is untrue.
The larger picture: UK benches are also considering AI
A tax tribunal judge publicly revealed a limited use of artificial intelligence (AI) to help prepare a disclosure finding at about the same time, according to UK legal media. This appears to be the first published UK judicial decision to make such a claim. The distinction is crucial because the judge maintained control and accountability over the decision while using AI as a tool rather than a replacement. This demonstrates the asymmetry: whereas judges may experiment under rigorous self-imposed boundaries, parties that contract out fact-finding to an unreliable model court are doomed.
When AI creates legal work products, what goes wrong?
When it comes to controversial trade mark work, there are three common failure modes:
1. Hallucinated authorities: references to cases that never happened or passages that were misquoted. Credibility is quickly damaged by these.
2. Overly generalised principles: AI tends to make bold generalisations (such as “post-sale confusion always matters”) without taking into account the subtle boundaries set by previous case law. Your foothold slips when the other side points to the true ratio.
3. Evidence mismatch: The story produced by AI outweighs the supporting data (for example, claiming “wide UK reputation” with scant evidence). Rhetorical exaggeration is not tolerated by the Appointed Person.
A useful guide for IPEC and UKIPO practice
1) Consider AI a junior assistant rather than a co-counsel.
Use it simply to speed up your own analysis if you’re using it for initial drafts or summaries. Check all references and information against the original source (the exhibits, witness statement, market data, register extract, and tribunal ruling).
2) Re-anchor arguments in UKIPO practice following SkyKick.
Boilerplate recycling is a favourite of AI. However, the UKIPO has been more critical of pleas in bad faith and too broad specifications following the Supreme Court’s SkyKick ruling. Commercial justification and targeted specifications are emphasised in the Office’s July Practice Amendment Notice (PAN 1/25) and later rulings. Refit your AI-drafted argument to the most recent guidelines if it is based on out-of-date pre-2025 generalisations.
https://www.gov.uk/government/publications/practice-amendment-notice-125
3) Adjectives come after the evidence.
What is the true purpose, who is perplexed, where, when, how many sales, which channels, and what advertising are some of the questions that tribunals will examine. Before allowing a model to “polish” the tale, create the case file, which should include market share, bills, web analytics, and surveys.
Don’t be lazy use it as a tool nothing more.
4) Maintain a “AI audit trail”
Keep your suggestions, iterations, and—most importantly—your manual verification notes if a tool helped with the drafting. Even though you might never need them, you’ll be happy to have them in case the tribunal looks into the matter.
5) Develop teams rather than just tools.
Update internal precedents using the most recent information on variant/promotional usage, partial revocation, and Appointed Person practice points from the Court of Appeal’s easyGroup rulings, as well as post-SkyKick case law. The greatest remedy for irresponsible prompt-and-paste behaviours is this.
A lesson for brand owners
Safe use of AI is the competitive edge, not “using AI.” You maintain narrative control and steer clear of preventable credibility strikes when you combine methodical evidence collection with meticulously checked drafting. AI is a sharp tool that needs to be treated carefully in UK trade mark appeals, as the Appointed Person’s late-summer caution makes clear.
Don’t be lazy. You have come this far, in your career; don’t let some software make a fool of you.
