In May I wrote for Wonkhe about two reports received through an English university’s reporting platform: a student reporting a coursemate’s public Instagram reels about faith and sexuality, and two students on a Middle East politics unit reporting each other over comments made in seminars. I suggested that the guidance gave the people handling those reports little practical help, and I asked for three things: an acknowledgement that student-on-student reports are a distinct scenario, a mapping of expectations across condition E6 (harassment and sexual misconduct) and the free speech guidance, and worked examples of triage decisions.
What changed in August
On 5 August 2026 OfS amended Regulatory Advice 24, alongside the final rules for the free speech complaints scheme that opens on 1 September. The press release describes “minor updates”. Jim Dickinson has set out on Wonkhe why the changes amount to more than that: the three-step framework has been revised to follow the High Court’s April judgment in University of Sussex v Office for Students, so that the same facts can now count at step 2 and step 3; new paragraphs cover Article 17 of the European Convention and a provider’s own obligations under the Human Rights Act; four examples were amended and one added; and paragraph 124 still opens with the old sequencing, which the rest of the document has now revised. His conclusion, which I share, is that the framework has been corrected while most of the examples have not been reassessed against it.
The examples are the most specific part of the guidance, and most of them were unchanged by the update. So I decided to analyse all 55 in the amended version: the 54 carried over from June 2025, of which four (35, 38, 40, and 47) were amended in August, plus new example 24A.
Each was coded for area of activity, actors, outcome, strength of conclusion, practical depth, and the interest or viewpoint at stake. Four patterns are of interest for those of us who handle student conduct reports.
Patterns across the 55 examples
Firm and qualified language are not evenly distributed.
Where an example finds that an institution breached the duty by restricting speech, the language is firm and directive: the recurring formula is “It is likely that University B has breached its ‘secure’ duty”, and example 20’s institution “should now take the reasonably practicable step of renewing Professor X’s invitation”.
Where an example concludes that a restriction or intervention is acceptable, the language is qualified, sometimes with two or more conditions attached; amended example 35 uses “may” six times before its conclusion, “may be unlikely to be in breach”.
Of the eight examples with the most qualifying language, six permit a restriction.
The likely explanation for this is structural rather than editorial: the statutory duty is to secure speech, not to restrict it, so telling providers what they must not restrict is safer for a regulator than pre-approving a restriction.
However, the practical effect is still unbalanced, because the decisions that university conduct teams face are almost all in the qualified category.
The qualification follows a consistent order.
Where the restricted speech is unlawful, the conclusions are firm.
Where a restriction protects an institutional function, safety, or professional standards, they are moderately qualified.
Where a restriction protects a person’s dignity from lawful but harmful speech, the qualifying language is the most extensive in the set: the gay student in example 4, the trans student in amended example 35, and the faith group in example 47.
Example 35 is the clearest single illustration because both sides of one dispute now appear in the same example: the blanket pronoun rule receives a firm finding of likely breach, while the passage that would protect a trans student from targeted misgendering is the most qualified in the document. The effect is that the guidance is firmest about the situations conduct teams see the least, and most qualified about the ones they are likely to see the most.
Most of the practical detail tells the institution to reverse something.
Remove a clause, amend a policy, reinstate a member of staff, renew an invitation.
Content a practitioner could sensibly reuse is scarce; model wording appears in four examples, multi-step remedies in six, and a worked, step-labelled method in exactly one, new example 24A.
The only two examples that describe the steps of an investigation, 9 and 10, are taken from the E6 guidance rather than written for this document. The set is specific about removing restrictive wording and policies and says little about how to run a lawful protective process.
Students appear in the set in four roles.
Most often they are the source of the complaints, petitions, and protests that the provider is expected to resist; ten examples take that form. Occasionally they are protected speakers whose own expression the institution has restricted. They are protected from other people’s speech firmly where that speech is unlawful, and cautiously where it is lawful but harmful. And exactly one example out of 55 involves one student reporting another: example 1, which stipulates that the reported conduct is probably unlawful harassment.
The question that makes these reports difficult in practice (deciding whether reported speech is lawful and what follows if it is) is settled by assumption in the only example that raises it.
What the amendments did to the examples
All five changes relate to when a restriction or intervention is acceptable, and all five add reasoning rather than certainty.
Example 24A models a step-by-step analysis of a professional placement restriction; example 35 adds an Article 8 analysis for targeted conduct; example 38 adds a least-intrusive-means point; and example 47 now acknowledges that relocating an event interferes with speech, then justifies the interference by balancing the Convention rights involved.
I do think these are improvements. Example 47 is the first time the set shows what a documented balancing exercise involves, and the Article 8 reasoning in example 35 and new paragraph 64A is the first content in the guidance that a decision-maker acting to protect someone can use directly. The remaining 50 examples were not reassessed against the revised framework, so their outcomes are a better guide than their reasoning.
The fifth change is the one those of us in student casework may want to read most carefully: in the June 2025 version of example 40, a university investigating a professor’s protest speech was at fault because the investigation itself punished lawful expression. In the August version that sentence is removed, and University B’s failing is that it “did not have an adequate triage process”, and “a rapid triage process” is a reasonably practicable step it “should now take”.
The regulatory concern is now the quality of the triage that precedes an investigation, rather than the existence of one…
Triage is named but not described
That change makes triage the decisive stage of a compliant complaints process; it is also a process the document does not describe.
The investigation paragraphs, which were not amended in August, still refer to “a preliminary assessment/triage to assess whether to commence an investigation” without saying what that assessment involves. Example 40 now names an inadequate triage process as the defect and a rapid one as the remedy, again without saying what an adequate one contains.
None of the 55 examples, old, amended, or new, outlines a triage: what is assessed, in what order, against which questions, by whom, on what documentation, or what is said to a reporting student when the decision is not to proceed.
That brings me back to the two reports from my piece in May. Example 24A is about a medical student, and it is the only example in the set with labelled step-by-step reasoning. But it models an institution restricting a student’s speech on placement to meet professional standards; it does not address what to do when the concern arrives as a report from another student (a reasonably common scenario).
A conduct officer holding the Instagram report, or the two mutual reports from the politics seminar, is in much the same position as when the guidance first appeared: a firm instruction that lawful speech must not be punished for its viewpoint; a permission to act where dignity is at issue, subject to multiple conditions; a new instruction to triage rapidly; and no account of what a compliant triage of either report involves.
What the guidance still doesn’t tell us
Measured against what I asked for in May, the update addresses the third point in part and neither of the others.
Student-on-student reports are still not acknowledged as a distinct scenario; the relationship with E6 still has to be assembled from cross-references; and triage is now more prominent in the guidance without being any better described.
The summary of the updated guidance says OfS expects “to publish further examples in the future to reflect experience across the sector”. From 1 September that experience will include a complaints scheme through which staff and visiting speakers can bring free speech complaints directly to the regulator, while students’ complaints continue through the OIA, with the registration condition following in April 2027.
Further examples are welcome; worked triage decisions in student-on-student cases would be the most useful to publish first.