The Use of AI in Disclosure – guidance in the Competition and Appeals Tribunal

On 16 December 2025, the Competition Appeals Tribunal (“the CAT”) ruled on how the parties are to use AI in the disclosure exercise in Gormsen v Meta Platforms Inc [2025] CAT 85. The ruling provides a helpful overview of the approach to disclosure in the CAT under its procedural rules, and also provides guidance for similar large and complex exercises in the future, potentially nodding to other large class actions ongoing in England.  However, the factors identified may also be of interest to those considering use of AI in other disclosure proceedings such as High Court Litigation, especially as one of those sitting on the tribunal is an author of a leading textbook on disclosure.

The collective proceedings, brought by Dr Liza Lovdahl Gormsen on behalf of an estimated 46.6 million UK Facebook users, alleges that Meta abused its dominant position by imposing an unfair bargain, requiring users to permit extensive collection and use of Off‑Facebook Data as a condition of accessing Facebook. 

In its December ruling, the CAT issued general guidance on disclosure and characterised the use of technology in the proceedings as both “a curse and a cure.” The curse being the explosion of documents and data that may require disclosure, and a potential cure that could improve speed and reduce costs where used appropriately. 

The ruling quotes expert evidence extensively which explain the potential use of technology to assist review, namely technology assisted review (TAR) or continuous active learning (CAL) or, alternatively, an AI tool based on large language models (LLM). Here, we focus on the key takeaways regarding disclosure. 

Use of AI

• AI permitted but not mandated: The Tribunal was ‘not willing to be prescriptive’ and require the use of any particular tool or AI. Rather, the responsibility for this falls to the Designated Solicitor who should monitor and review the use of technology and adapt accordingly.

• ‘AI intensive document review tools’ do not appear ‘generally inappropriate’ in circumstances where stringent human tests and checks are in place. Although, the Tribunal recognised concerns raised by Meta regarding whether LLM based AI tools can scale to the size and complexity of disclosure in these proceedings.

• AI in document review not considered ‘experimental’ when supplementing human review: The Tribunal noted that there ‘are likely to be many cases in which utilising AI technology in document review will achieve significant reductions in time and cost of review.’

• Use of AI for document review may change the analysis for what is a proportionate search: The Tribunal recognised that AI may have altered what is deemed ‘reasonably necessary and proportionate’ given the reduction to time and cost burdens. However, they also recognised that other factors need to be considered, noting that not all documents are amenable to AI led review and the tools themselves require significant financial and human resources.

The High court endorsed the use of predictive coding, a term that includes TAR and CAL, in document review in Pyrrho v MWB [2016] EWHC 256 (Ch). In that case, both parties agreed that predictive coding could be used. The issue was revisited in Brown v BCA Trading [2016] EWHC 1464 (Ch), where the Court made an order for predictive coding in disclosure, even where both parties did not agree on There is yet to be a similar judgment on whether an LLM based AI approach can be used in document review. 

Whilst the CAT rulings are not binding in the High Court, the Tribunal’s comments, based on expert evidence, will likely be of note.  It is notable that a number of the points are similar to factors considered in Pyrrho: consideration of what is a reasonable and proportionate exercise; costs; time; the accuracy and consistency of human reviewers; and industry practice.  

The Disclosure process

The ruling is also of note for a few other reasons.

First, the Tribunal recognised the potential that it would order the disclosure of the Disclosure Protocol that provides instructions to those performing the document review:

The Tribunal is not requiring Meta at this stage to disclose the Disclosure Protocol, but if further down the line the Tribunal considers that it requires sight of it, then Meta should be ready to disclose it as part of the Tribunal's case management of disclosure.

Second, the Tribunal recognised the risk of inconsistent approaches where there are large document review teams, albeit that these risks could be mitigated, such as  through disclosure protocols and quality control.

Experience tells the Tribunal that the larger and more complex the disclosure exercise, and the larger the team working on disclosure, the more likely that errors will be made and inconsistencies in searches arise.

Third, the Tribunal noted the importance of how disclosure workstreams are structured, including through the use of a ‘Designated Solicitor’ for disclosure:

In other cases, it has been found to be efficient and give the best outcome if the Designated Solicitor is not working as part of the case team but is simply concentrating on the disclosure exercise. It has the advantage of giving a degree of independence and ability to push back on what members of the case team are saying – it is desirable to have someone who is going to exercise independent judgement and with the strength and ability to contest that certain disclosure should not be provided.

For more information about the law, technology and practice of disclosure, contact Tom Whittaker.

This article was written by Ellie Mills.