Summary

  • ComReg’s 8 October release describes guidance for AI use in disputes involving consumers, business users and providers in phone, broadband and postal services.
  • The regulator recognises possible gains in clarity, organisation and accessibility, but says the party filing a submission remains responsible for checking its accuracy and relevance.
  • ComReg may ask for a different or shorter format and clarification about AI use in appropriate cases; the release does not set out a universal disclosure form or automatic penalty.

An AI tool can make a complaint easier to read and still make it less reliable. That tension is the point of ComReg’s new guidance: assistance with drafting is allowed, but the record sent into a dispute process remains the responsibility of the person or business that files it.

ComReg’s consumer-news release, dated 8 October, covers end-user disputes involving consumers and business users of phone, broadband and postal services, as well as providers that are parties to those disputes. The regulator says AI may help prepare submissions, organise information, improve clarity and overcome language or accessibility barriers. It also lists familiar failure modes: incomplete or fabricated material, references that cannot be verified, inaccurate facts or citations, and submissions so long that they slow resolution.

The publication listing identifies the guidance as Information Notice ComReg 26/66, posted on 6 October. The two dates matter: the notice record predates the public news release, but neither page describes a blanket ban on AI-assisted submissions. ComReg instead says use should be transparent and that AI must not replace human judgment. Users remain responsible for accuracy; they should check facts and references and make sure the material is relevant rather than relying on generated text without human review.

That allocation puts the verification step at the point where a document becomes evidence in a case. A drafting system may produce fluent prose, but fluency does not establish that a cited rule exists, that a date is right or that a summary reflects the underlying correspondence. ComReg says inaccurate AI-generated content may delay a dispute or be treated as an abuse of process. It does not say that every error will be treated that way.

The regulator retains a case-level response. In appropriate cases it may request that information be resubmitted in a different or shorter format and ask how AI was used to prepare it. The release does not specify a standard disclosure form, a threshold for such requests or a fixed response to an error. That leaves room to address a defective submission without making the mere use of an AI tool disqualifying.

The access trade-off deserves attention. Help with language and accessibility can make a process more usable; additional checking demands can also fall hardest on people with the fewest resources to verify legal references or reconstruct a long record. ComReg’s statement that AI can help overcome those barriers sits alongside its insistence on human review. Whether those aims work together will depend on how requests for clarification are used in practice, not on the guidance’s endorsement of a tool alone.

For now, the evidence is limited to ComReg’s release and publication listing. They do not show how often AI is used in these cases, what disclosure looks like, how requests will be triggered, or whether consumers, business users and providers encounter the same process. The governance question is therefore practical: can parties use assistance to make a case clearer while keeping the underlying facts checkable by the party and the regulator?

Primary sources