The Appointed Person has warned litigants against the misuse of generative AI (GenAI) in crafting legal submissions, following an appeal in trade mark opposition proceedings in which it had been used carelessly by both parties (Pro Health Solutions ltd v ProHealth Inc., BL O/0559/25, 20 June 2025).
Facts
The appeal arose out of a decision by the Registrar to uphold an opposition, by Pro Health Inc. (the respondent), to the registration by Pro Health Solutions (the appellant) of a word mark for a series of two words, PROHEALTH and Prohealth, in class 5.
The respondent successfully opposed registration under section 5(4)(a) of the Trade Marks Act 1994 (TMA), by proving that it had protectable goodwill through its sales and other activity under certain signs in the UK before the appellant’s application was filed. The Appointed Person (AP) upheld the Registrar’s finding that sales, even in small numbers, could suffice to establish protectable goodwill in the signs under which they were sold.
Misstatements of case?
However, before deciding on the substantive appeal, the AP made various observations relating to the appellant’s grounds of appeal of the appellant, who was unrepresented, and to both parties’ skeleton arguments.
The appellant’s grounds of appeal had cited cases which were all real, but gave three purported quotes from those decisions which did not exist. His skeleton argument included six cases, two of which had erroneous citations and three had summaries which substantially misrepresented the relevant case. In addition, the skeleton mostly comprised factual issues which were not relevant to the issues in this case, as well as unhelpful arguments.
The appellant, when questioned on these by the AP during the hearing, admitted and apologised that he had drafted these documents with the assistance of ChatGPT.
In addition, the respondent’s representative, a trade mark attorney, had relied on three cases in his skeleton argument which, while genuine and correctly cited, did not stand for the propositions asserted. The representative failed to find the relevant paragraphs during the hearing, and then claimed the references came from two practitioners’ books, but the AP was unable to find these propositions in either book.
Fabricated submissions
The AP made extensive comments on the parties’ conduct in relation to their use of GenAI, and warned that:
- Even if unrepresented parties are given greater latitude in conducting their case, they are still under by a duty not to mislead the civil courts, the registrar or AP.
- While unrepresented litigants should not be punished for honestly advancing arguments that are irrelevant or based on genuine misunderstandings of the law, there is a difference between this and relying on laws which have been fabricated.
- Fake citations can involve making up a case entirely, making up quotes from a real case, and also making up legal propositions and attributing them to real cases that are not relevant to those propositions.
- It does not matter whether fabrication took place with or without the aid of generative AI.
Sanctions, contempt and costs
Neither the AP nor the registrar had the same power as the civil courts to impose sanctions, make a wasted costs order, or deal with contempt of court summarily (although the AP emphasised that none of the conduct here came close to contempt).
The AP observed that where a party has acted unreasonably, including by relying on fabricated citations, costs outside the relevant scale/range (“off-scale” costs) should generally be awarded as a starting point. Such conduct could also breach a lawyer’s regulatory duties (including those of a trade mark attorney), for which a professional representative could be publicly admonished or referred to their regulator.
In this case, the AP decided that:
- The unrepresented appellant was inexperienced and had apologised immediately and unreservedly, which weighed strongly in his favour. The AP did not order off-scale costs, as he went on to make no order as to costs due to the respondent’s conduct.
- The respondent’s representative’s actions were of great concern, due to his inability to explain where he sourced legal propositions, his apparent belief that he would not be expected to expand on points made in a skeleton argument, and that it was acceptable to use out-of-date textbooks. The AP decided not to refer him to the regulator, but warned him to “seriously reflect on how he conducts his practice in the future”.
IPO practice on using GenAI
The AP concluded his comments by offering helpful suggestions to the IPO on how to caution parties against relying on GenAI to craft submissions.
He noted that many unrepresented parties will know little about trade mark law, and may think that anything GenAI creates “will be better than they can produce themselves”. This meant a very clear warning needed to be given, “to make even the most nervous litigant aware of the risks they are taking”.
He suggested that the registrar considers adopting a practice, for both represented and unrepresented litigants, of including a very clear warning:
- in any correspondence requesting submissions, skeletons or other documents from the parties;
- which sets out explicitly the risks of using AI for legal research, drafting skeletons or written submissions; and
- setting out the professional consequences of putting fabricated citations before the registrar or Appointed Person.
Key takeaways
While we can expect further guidance for legal representatives to emerge through a recently-established Civil Justice Council working group, the AP’s warnings on the misuse of GenAI are similar to those issued to judicial office-holders themselves in Judiciary guidance, the latest version of which was published in May 2025.
The lesson from that guidance, and the AP’s guidance in this case, is clear. While parties and judges alike are not prevented from using GenAI tools in the conduct of their cases, they should clearly understand the risks involved, and the potential consequences if they are caught misusing GenAI in their work.