Misconceptions

What the law requires

“The Supreme Court settled it in For Women Scotland, so we can exclude trans people from single-sex provision.”

It settled what “sex”, “man” and “woman” mean in the Equality Act. It created no duty to exclude anybody. The exceptions that permit single-sex provision sit in Schedule 3 and are permissive, never mandatory, and the regulator’s 2026 Code still treats a service open to all as the ordinary starting point.

What is actually the case

The expectation is reasonable. This was the highest court in the land settling something that had been argued for years, and it landed in a period when organisations were waiting for somebody to tell them what to do. A judgment that sounds definitive is very attractive to an organisation that wants the question closed. The case is also almost always reported in a single sentence, which hides what it was actually about: Scottish legislation on the representation of women on public boards, and whether a Gender Recognition Certificate made someone a “woman” for the purposes of that measure.

What the Court answered was a question of statutory interpretation — whether those words in the Equality Act include certificated sex or biological sex. It held biological sex. That genuinely changes how the Act’s sex-based provisions operate, and it is settled. What it did not do is touch gender reassignment, which is a separate protected characteristic in the same Act and is unaffected. Nor did it create an obligation. The exceptions in Schedule 3, Part 7 permit a provider to depart from mixed provision where it can justify doing so. They do not require it. Reading a permission as an instruction is the most expensive single error in this area, because it converts a decision you would have to justify into one you believe you have no choice about — and then nobody writes down the reasoning, because nobody thinks there was a choice to reason about.

So the judgment did not move your default. If your provision was open to all before it, nothing in the judgment closed it. The question in each case is still whether a restriction is a proportionate means of achieving a legitimate aim, and the burden of showing that sits with the provider, not with the person seeking access.

What is genuinely unsettled

  • How the exceptions apply to particular configurations of premises and services is still working through. The EHRC’s 2026 Services Code came into force on 5 August 2026 and its application has not yet been widely tested.

Where this bites

If you changed a policy because you believed the judgment obliged you to, the reasoning behind that change was never recorded — because nobody thought there was a choice to record.

Read the detail

“Protection only applies once someone has had surgery, or at least a diagnosis.”

No medical step is required at any point. Section 7 of the Equality Act protects anyone proposing to undergo, undergoing, or having undergone a process to reassign their sex — and “proposing” means protection starts at the conversation, before anything else happens.

What is actually the case

This one is plausible because almost every other threshold in an employer’s life is evidenced by a document. Sick pay needs a fit note, a right to work needs a passport, an adjustment often follows an occupational health report. A characteristic that attaches on the strength of what somebody tells you, with nothing to file, runs against the grain of how HR is trained to work. The word “reassignment” does not help either — it sounds surgical, and it is not.

Section 7 turns on a process, not a treatment, and it explicitly includes proposing to undergo one. Changing a name, pronouns or presentation can be the whole of it. No certificate is needed, and a Gender Recognition Certificate is a separate question in a separate statute. The practical consequence is the one organisations miss: protection begins at the moment somebody tells a manager they intend to transition. Everything that happens after that conversation — how it was received, who it was passed to, what changed in the rota — is inside the protected period, including any decision made while the organisation was still working out what its policy ought to be.

There is a second half to this that matters more in practice than the threshold does. Because protection needs no evidence, asking for evidence is itself the risk. A request for a certificate, a letter or a medical history before a name will be changed on an email account is a request for special category data you have no lawful need for, and it is the kind of exchange that reads badly in a tribunal bundle years later.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

The protected period starts at the first conversation, not when your policy catches up — so the decisions you make while you are still deciding are already inside it.

Read the detail

“A Gender Recognition Certificate settles it. / A GRC is meaningless now.”

Both are wrong, and they are the same mistake in opposite directions. Section 9 of the Gender Recognition Act makes the acquired gender the person’s gender for all purposes unless a specific provision says otherwise — and the Equality Act’s use of “sex” is now one of those provisions. The certificate is fully valid; it simply is not doing the work that particular word is doing.

What is actually the case

Both readings are understandable, and they tend to arrive in sequence. Before 2025, “for all purposes” in section 9 sounded comprehensive, because it reads comprehensively. After For Women Scotland, the conclusion that the certificate had been hollowed out sounded equally obvious. Each reading takes a holding about one statute and applies it to the whole of a person’s legal position.

What section 9 actually does is create a presumption with an exception clause built into it: the acquired gender applies for all purposes unless a specific provision of law says otherwise. For Women Scotland identified the Equality Act’s use of “sex”, including its single-sex exceptions, as somewhere the presumption does not reach. Outside that, the certificate continues to do everything it did — records, marriage, pensions, dealings with public authorities. It was not weakened; the boundary of one of its effects was drawn.

The part that most often catches an employer out is not access at all. Section 22 of the Gender Recognition Act makes it an offence to disclose protected information about a person’s gender history acquired in an official capacity. For most organisations that is the GRC’s live operational effect: a confidentiality duty sitting on top of the ordinary data-protection regime, breached most often by a well-meaning handover — a note to a new manager “for context”, a mention to a colleague who is about to work with them.

What is genuinely unsettled

  • Which other statutory regimes count as a “specific provision” displacing section 9 has not been exhaustively mapped. It depends on which statute governs the situation in front of you, which is why a decision turning on legal sex specifically is one to take advice on rather than answer from a template.

Where this bites

If a decision turns on legal sex, the question is which statute governs it — not whether a certificate exists. And whatever you decide, the confidentiality duty applies either way.

Read the detail

“Nonbinary people obviously are not covered — the protection is about transitioning from one sex to the other.”

That confidence is misplaced, but so is confidence the other way. A tribunal in Taylor v Jaguar Land Rover held a gender-fluid and nonbinary employee protected under section 7; a later tribunal in Lockwood v Cheshire and Wirral took a narrower view after For Women Scotland. Both are first instance, neither binds, and the position is genuinely unsettled.

What is actually the case

The claim reads naturally off the statute. Section 7 speaks of a process to reassign “sex”, and sex in ordinary use has two values, so a person who identifies outside both can look like someone the wording was not written for. The drafting is from 2010 and the vocabulary has moved since.

Two tribunals have now looked at it and come to different places. In Taylor v Jaguar Land Rover a gender-fluid and nonbinary employee was held to be within section 7, and the tribunal found both harassment and direct discrimination. In Lockwood v Cheshire and Wirral NHS Foundation Trust, decided after For Women Scotland, a tribunal held that a nonbinary person with no intention of transitioning did not have the protected characteristic. Both are first-instance decisions. Neither binds another tribunal, and no appellate court has resolved the tension between them. Anyone telling you this question is settled — in either direction — is telling you something the case law does not currently support.

What that means operationally is narrower and more useful than the legal question. Almost none of what an organisation actually does here depends on the answer. Harassment on grounds of perceived gender reassignment is covered regardless. So is harassment related to sex. Dignity at work, data protection, and the ordinary duty not to run a hostile environment do not wait for a tribunal to classify anybody. An organisation that treats the coverage question as the thing to resolve before it acts has picked the one part of this that nobody can currently answer, and made it a precondition for the parts that are clear.

What is genuinely unsettled

  • Whether, and in what circumstances, a nonbinary or gender-fluid person holds the protected characteristic. Taylor and Lockwood point in different directions, both are first instance, and it will take an appellate decision to settle it.

Where this bites

Waiting for the coverage question to be answered before you act means waiting on the one part of this nobody can answer — while the harassment and data duties that do not depend on it are already running.

Read the detail

“This is employment law — it does not apply to how we run our services.”

The Equality Act binds duty-bearers in several defined roles: employers, service providers, public authorities and associations. Each is a different regime with different tests, and an organisation is frequently more than one of them at once.

What is actually the case

The belief comes from where most people meet this law. Training is usually run by HR, the cases that make the news are usually tribunal claims, and the internal owner is usually an employment specialist. If every encounter you have had with the Act came through an employment frame, an employment-shaped understanding is what you end up with.

The Act places duties on specific actors in specific roles. Employers are one. Service providers are another, under a different part of the Act with its own exceptions. Public authorities carry an additional duty to have due regard to eliminating discrimination and advancing equality of opportunity — a duty about process, not about reaching a particular outcome. Associations are a fourth. A hospital is an employer and a service provider and a public authority in the same corridor, and the analysis is not the same in each role. A decision that is defensible as an employment matter can fail as a services matter, because the exception you were relying on lives in the wrong part of the Act.

There is a converse worth stating, because organisations get it backwards in the other direction too. The Act does not regulate what individuals privately think or believe. A colleague holding a view is not a duty-bearer and is not breaching anything by holding it. The reason their conduct at work can still become the organisation’s problem is structural: the Act makes an employer liable for what its staff do in the course of their work. In Unite the Union v Nailard the point was made sharply — a policy in the handbook does not, on its own, discharge that. What answers the liability is evidence: training delivered, monitoring done, action taken when something went wrong.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

Before you reach for an exception, establish which role you are acting in. The exception you are relying on may sit in a part of the Act that does not govern what you are doing.

Read the detail

“It is a single-sex service, so we have to exclude trans people.”

An exception you may use is not a duty you must discharge. Schedule 3 of the Equality Act permits a departure from mixed provision where it can be justified; it never requires one. The 2026 Services Code treats provision open to all as the ordinary starting point, and puts the burden of justifying a restriction on the provider.

What is actually the case

Two real things feed this. The first is that “single-sex service” sounds like a category with rules attached, in the way that “licensed premises” or “regulated activity” does — so it is natural to assume that once the label applies, the consequences follow automatically. The second is that many services which describe themselves as single-sex are not, in law, single-sex at all. They are mixed services with a women-centred focus, which is frequently the better design and a great deal easier to defend.

Where the exceptions do apply, they are permissive. Schedule 3, Part 7 sets out the circumstances in which a provider may restrict a service by sex. Nothing in it obliges anybody, and nothing in For Women Scotland converted it into an obligation — the judgment clarified a definition, it did not move the default. The options are also a spectrum rather than a switch: open to all, separate provision side by side, or a restriction, each with a different justification burden. Most “should this be single-sex?” questions are answered further up that list than the person asking first assumes.

The sequence matters more than the label. Is this service single-sex? Does it need to be? What other options exist? And how would we operate and maintain the restriction — including the cost of policing it, which is where most schemes quietly fail. No UK document reliably evidences biological sex at a service point, appearance evidences nothing, and challenging someone at a reception desk is itself a dignity and data-protection exposure, because sex and gender reassignment are both special category data.

What is genuinely unsettled

  • How the exceptions apply to specific service configurations after For Women Scotland and the 2026 Code is still developing, and has not been tested widely.
  • What amounts to adequate operational management of a restriction — the policing question — has no general answer and is decided on the facts.

Where this bites

The decision is not whether the label applies. It is whether you can justify a restriction, and whether you can operate it without creating a second exposure at the point of challenge.

Read the detail

“Only someone with the protected characteristic can bring a claim about it.”

Not since Rollett v British Airways. The Employment Appeal Tribunal held in 2024 that a person can bring an indirect discrimination claim without holding the characteristic themselves, provided the rule leaves them worse off in the same way. It is an EAT decision, not settled higher-court law.

What is actually the case

This was the orthodox understanding, which is why it is still repeated by people who learned it correctly. Indirect discrimination was taken to require the claimant to share the characteristic connected to the disadvantage, and risk assessments were built on that assumption — count the people in the group, size the exposure accordingly.

Rollett tested the assumption directly and the Employment Appeal Tribunal rejected it as a fixed rule. What matters is the disadvantage, not the claimant’s own characteristics: if a rule puts a protected group at a particular disadvantage, and it leaves you worse off in the same way, you may be able to challenge it whether or not you are in that group. A blanket rule therefore no longer has to disadvantage only the protected group to be challengeable, and the pool of people who can challenge one badly designed policy is wider than it was.

The practical consequence is about design rather than headcount. A blanket rule is now exposed from more directions at once, which is a direct argument for case-by-case, evidenced decisions over categorical ones — and a direct argument for the impact assessment, because that is the document that shows you identified who a rule affects before you applied it to everybody. An organisation that sized its risk by counting the members of a protected group has sized it on the wrong number.

What is genuinely unsettled

  • Rollett is an Employment Appeal Tribunal decision. It binds tribunals but has not been tested in the Court of Appeal or above, so the boundaries of the “same disadvantage” reasoning are not fixed.

Where this bites

If you sized the risk of a blanket rule by counting how many people hold the characteristic, the number was wrong — anyone caught by the same disadvantage may be able to challenge it.

Read the detail

“If we include, we get sued. If we hold the line, we do not.”

Claims have succeeded in both directions, and the ones that failed failed on evidence rather than on which way the organisation went. What distinguishes the outcomes is not the direction of the decision but whether anyone owned it, weighed the alternatives and wrote down why.

What is actually the case

The asymmetry feels real because the two kinds of risk are visible in different ways. A claim brought against an inclusive policy arrives as a named case with a hearing and coverage. The risk of not acting arrives as a grievance that was never logged, a resignation, a manager improvising in a corridor — slower, quieter, and rarely traced back to the decision that caused it. One is a headline and the other is attrition, so the first feels like the whole of the exposure.

The record does not support the asymmetry. Facilities claims brought by women objecting to trans-inclusive arrangements have both succeeded and failed. Harassment and discrimination claims by trans employees have done the same. Where a claim failed, it failed because the arrangement had no practical effect on the claimant, or because causation was not established, or because the statutory threshold was not met — not because the employer had picked the safe direction. And in at least one case the employer won while still being criticised for having run no consultation and produced no impact assessment: the claim failed on its facts, and the process was found wanting anyway.

What actually separates the outcomes is ownership and record. Our own survey of 136 UK organisations found 53.7% with nobody accountable at senior level for this area, and 6.1% tying inclusion outcomes to any executive measure. A risk with no owner has no escalation path, so it matures quietly until it is a crisis. These decisions rarely come apart at the point of judgement. They come apart because the risk the decision created was never anybody’s to carry.

What is genuinely unsettled

  • Most of the facilities decisions cited in either direction are first-instance tribunal judgments. They bind nobody, and an appellate decision could move the balance of exposure in ways none of the current cases predicts.

Where this bites

There is no direction you can pick that removes the exposure. There is only a decision you can evidence, or one you cannot.

Read the detail

Misconceptions

Process and evidence

“We are not allowed to ask anything about this — data protection forbids it.”

Data protection does not ask whether you may. It asks what for, on what basis, and for how long. Gender-history information is not itself named in Article 9’s closed list, but it usually qualifies because it reveals or is bound up with data concerning health — so the safe assumption is to handle it as special category, which means a lawful basis and a separate condition before you collect it. “Just in case” is neither.

What is actually the case

The belief is a reasonable response to real complexity. Special-category data carries a genuinely higher threshold, the penalties are real, and most people meet the regime through mandatory training that emphasises restraint. Faced with a rule you half-remember and a topic that feels charged, “do not touch it” is the safe-sounding conclusion. It is also the conclusion that produces the worst outcomes, because it converts a question about purpose into a blanket refusal.

The regime is a discipline, not a prohibition, and it is worth being precise about what triggers it. Gender-history information is not in the closed list of special categories at Article 9; it usually falls there in practice because it reveals, or sits close to, data concerning health, which is. That distinction matters to a controller deciding whether it needs an Article 9 condition at all — but the protective default is to assume it does, because the alternative is deciding case by case whether this particular record happens to reveal health, and getting that wrong is expensive. So: a lawful basis for processing, a separate condition for the special-category element, and then the ordinary obligations follow: collect only what you need, use it only for what you collected it for, restrict who can see it, decide how long you keep it, and tell the person. What that rules out is speculative collection. Holding gender-history data because it might be useful one day has no basis, and the absence of that data in your records is not a gap to be closed.

One thing genuinely does approach a prohibition, and it is not about asking. Section 22 of the Gender Recognition Act makes it a criminal offence to disclose that a person has applied for or obtained a certificate, where that was learned in an official capacity, and the exceptions are exhaustive. So the sharp edge is on the way out, not the way in — the handover note, the word to a new manager, the granular monitoring breakdown so small that one person is identifiable in it.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

The question is never “are we allowed to ask”. It is “what is this for, what is our basis, who will see it, and when does it get deleted” — and if you cannot answer those, do not collect it.

Read the detail

“We cannot act until we have a policy in place.”

A policy is not what makes a decision defensible, and waiting for one does not pause the duties you already have. What holds up under challenge is a documented, reasoned decision on the facts — which you can make today, policy or no policy.

What is actually the case

Writing a policy feels like doing the work, and that is exactly why it is such a comfortable place to stop. It is visible, it is finishable, and it produces something you can point at when somebody asks what you do about this. Our own research named the pattern the Paper Shield: a formal policy with none of the operational infrastructure to implement it — no manager guidance, no systems capability, no measurable outcome. Of 136 UK organisations surveyed, 36.4% were in precisely that position: a policy in place, and managers given nothing to apply it with. It is rarely dishonest. Most organisations with one believe it represents real commitment.

The gap only shows when it is tested. A tribunal or a regulator does not ask whether a policy exists; it asks what happened when the policy met a real case. A policy with no decision framework behind it is fragile by design, and the fragility is invisible until the moment it matters. Meanwhile the duty does not wait: protection begins when somebody tells a manager they intend to transition, and every decision taken while the organisation is still drafting is already inside the protected period.

The inversion worth taking away is that the policy is downstream of the capability, not upstream of it. An organisation that can reason a case properly — establish which regime governs, weigh the options, record why — can write a policy afterwards that describes what it actually does. An organisation that writes the policy first has usually described what it hopes it does, which is the document that fails under pressure.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

If somebody has already told you they are transitioning, the decision is in front of you now. Drafting is not a holding position; it is just an unrecorded decision.

Read the detail

“Until the law is clearer, the safest thing is to do nothing.”

Inaction is a decision with the same exposure as any other, minus the record that would defend it. The Equality Act has been in force since 2010, For Women Scotland was decided in 2025, and the EHRC Services Code came into force on 5 August 2026. There is no pending event to wait for.

What is actually the case

Waiting feels responsible, and in most of corporate life it often is. Hold off, let the position settle, act once on good information rather than twice on bad. The instinct is reinforced here by genuinely shifting ground: a Supreme Court judgment, a new statutory Code, tribunal decisions arriving in both directions. If ever a subject invited patience, this one looks like it.

But the waiting has no terminus, and the thing being waited for has already happened. The Code was the last scheduled event and it commenced on 5 August 2026, replacing the 2011 Code the same day. What remains uncertain is how the exceptions apply to particular facts, and that is not the kind of uncertainty that resolves — it is decided case by case, permanently. An organisation waiting for the point at which the answer is obvious is waiting for something the structure of the law does not produce.

Meanwhile the exposure accrues in a form nobody logs. A manager improvises in a corridor, a grievance goes unrecorded, someone resigns, and none of it gets traced back to the decision not to decide. The comparison that matters is not between acting and waiting; it is between a decision you can evidence and one that happened anyway without anybody owning it. Inaction produces the second kind, and produces it invisibly.

What is genuinely unsettled

  • How the 2026 Code applies to specific service and facilities configurations is developing and will be tested through cases rather than resolved in advance.

Where this bites

While the organisation waits, managers are still deciding — one situation at a time, without guidance, and without anything written down to defend them.

Read the detail

“We consulted before we changed the policy, so we have done what was needed.”

Consultation only helps if it reached the people the policy affects. In LS v NHS England the employer consulted the staff network that benefited from the policy and no others, could not produce its equality impact assessment, and lost — because a less discriminatory option existed and had not been taken.

What is actually the case

The belief is understandable because consultation is the step organisations have been trained to treat as the safeguard. It appears in every governance checklist, it generates minutes, and it feels procedurally sound. If you consulted, you followed the process.

LS v NHS England shows how that can still fail. The employer consulted its disability staff network before issuing a trans-inclusive facilities policy, but no women’s network and no faith network — neither existed at the time of drafting. So the policy was only ever tested against the interests of those it benefited. The tribunal treated the absence of a forum for an affected group as a finding in itself, not an excuse. Asked for the equality impact assessment behind the policy, the employer could not produce it — and “we could not locate it” is a sentence that decides cases.

The decisive point was narrower and sharper than consultation. Justification failed because a less discriminatory means of achieving the same aims existed and had not been taken: fewer than five staff identified as trans women out of somewhere between sixteen and nineteen and a half thousand, and directing that small number to gender-neutral facilities was available rather than reshaping single-sex provision for the many. Where a lesser measure exists and is not taken, the justification defence fails — however well-minuted the consultation was.

What is genuinely unsettled

  • LS is a first-instance employment tribunal decision on liability only, with remedy outstanding. It binds nobody, and the boundaries of the lesser-measure reasoning have not been tested on appeal.

Where this bites

Two questions decide it: did you consult anybody the policy disadvantages, and can you produce the assessment today if somebody asks for it.

Read the detail

“Better not to write the reasoning down — it only creates something disclosable.”

The opposite. Undocumented reasoning is functionally indistinguishable from no reasoning once a decision is challenged, and the record is what converts a contested judgement into a defensible one.

What is actually the case

The instinct has real roots. Documents do get disclosed, a clumsy sentence in a minute can be read back years later in a hearing, and most people have seen an email quoted against its author. Saying less looks like reducing surface area.

It reduces the wrong surface area. What is examined in a challenge is not whether a document existed but whether the organisation can show it reasoned. A defensible record shows six things as at the day of the decision: the question, the facts relied on, the legal basis, the options considered, the proportionality rationale applied to these particular facts, and a review date. Absent that, the organisation is left asserting that careful thinking happened, with nothing to distinguish it from a decision taken on instinct — and reasoning reconstructed after a challenge arrives reads exactly like reasoning reconstructed after a challenge arrives. LS v NHS England is the concrete version of the cost: asked to produce the equality impact assessment behind its facilities policy, the employer could not, and the tribunal treated that absence as a finding rather than an administrative slip.

The framing that makes this easier is that the record is not authorising a position, it is authorising a process. A decision recorded as a reasoned, proportionate, case-by-case process holds up whichever way the law subsequently moves. A decision recorded as a fixed line becomes fragile the moment the law or the facts shift under it — which is the actual risk in a developing area, and the one that silence does nothing to reduce.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

If it is challenged in two years, the only difference between a careful decision and a careless one will be what you wrote down at the time.

Read the detail

“Lived experience is anecdote, not evidence — we need to decide on the facts.”

It is evidence, of a specific thing: the impact of a system. What it is not is the outcome. Both halves of that matter, and organisations get it wrong in both directions.

What is actually the case

There is a real reason for the scepticism, and it is not hostility. Lived experience is usually delivered to organisations as a story — someone speaks at the front of a room, the audience is moved, the feedback forms say “powerful”, and then everyone goes back to work. The policy is unchanged, the manager facing a real situation is no better equipped, nobody was given responsibility for acting on what was heard, and there is no way of knowing whether it influenced a single later decision. Anyone who has sat through that is right to conclude something was missing.

What was missing was not rigour on the speaker’s part. Hearing a story is an activity, not an outcome. The failure is downstream: nothing was converted into a finding, an owner or a change. Treated properly, lived experience does something no dashboard can — it shows the consequences of arrangements that policies, metrics and formal complaint counts systematically fail to surface, because the people affected by them mostly do not complain. There is a figure for that: of 136 UK organisations surveyed, 30.6% estimated that fewer than one in ten of their staff would feel comfortable disclosing their identity. A complaint count is not a measure of impact in a population that is not disclosing.

So it is evidence of impact, and impact is exactly what a proportionality assessment has to weigh. That is also its limit. It does not decide the question, it does not override the legal analysis, and it does not settle what a proportionate arrangement is. An organisation that dismisses it loses the only visibility it has into effects it is legally required to consider. An organisation that treats the session as the outcome has held an event.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

Ask what finding came out of it, who owns that finding, and which decision changed. If there is no answer to all three, it was an event and not evidence.

Read the detail

Misconceptions

Facilities

“There is a settled answer on toilets and changing rooms — we just need someone to tell us what it is.”

There is no settled rule to be told. The statutory framework requires separate provision but does not resolve who uses which room, and an employment tribunal put the gap directly in LS v NHS England: there is no express legal right for a trans person to use the single-sex facilities matching their gender identity, and no rule requiring blanket exclusion either. What separated the recent cases was how the employer decided, not which group it favoured.

What is actually the case

The expectation is reasonable, and it comes from somewhere real. Parts of this area genuinely are rule-bound. The Workplace (Health, Safety and Welfare) Regulations 1992 require separate provision across Great Britain, and in England — England only — Approved Document T sets out toilet accommodation in some detail, while Scotland and Wales have no Part T equivalent. From there it is a short step to assuming there must be an equally clear rule about who may use which room. There is not one, and the cases organisations reach for do not supply it.

Four tribunal decisions are now cited constantly, and it is worth noticing that all four were brought by women objecting to a trans-inclusive facilities arrangement. They did not come out the same way. In Kelly v Leonardo every claim was dismissed: a genuine alternative existed and sat unused, the claimant could not show the arrangement had actually disadvantaged her, and the statutory thresholds were not met. In Hutchinson v County Durham and Darlington the claimants succeeded in part, with the tribunal reading “men” and “women” in the 1992 Regulations by reference to biological sex, and finding that a nominal alternative room is no alternative at all if it is markedly further away, of lower quality, or available to only some of the people who need it. In LS v NHS England indirect sex discrimination and harassment succeeded in part. Peggie v Fife turned substantially on how the employer handled the complaint once it arrived.

Croft v Royal Mail is the oldest and most quoted, and it is misread in a particular way. The Court of Appeal approved a phased approach in 2003 — interim arrangements moving to facilities matching the employee’s acquired gender as transition progressed. Its own most common misreading is treating “phased” as permission to delay indefinitely. The judgment’s words were “reasonable timescale” and “manage reasonably”, and both describe a process that moves. A phased approach with no next phase scheduled is not the approach Croft describes. Croft also predates the Gender Recognition Act 2004, the Equality Act 2010 and For Women Scotland; its reasoning remains influential but it is not the current governing test on its own.

Read together, the divergence is not about which group a tribunal preferred. It is about how the employer decided. LS failed its justification defence because a lesser measure existed and was not taken — fewer than five staff identified as trans women out of somewhere between sixteen and nineteen and a half thousand — and because the consultation ran past only the group the policy benefited, with no equality impact assessment the employer could later produce. Kelly survived because the arrangement had no practical effect and the alternative was real — though that employer was criticised anyway for having run no consultation and produced no equality impact assessment, which the tribunal described as best practice rather than compulsory for a private-sector employer. It states the through-line twice over: the claim failed on its facts, and the process was still found wanting. LS also supplies the sharpest point in this whole area: the same alternative facility, offered to different people, is a different act. Directing the few whose exclusion the law permits to a genuine alternative may be the proportionate lesser measure. Directing one of the many the provision exists to serve to that same room is displacement.

What is genuinely unsettled

  • Every one of these decisions except Croft is first instance. They bind nobody, LS is a liability judgment with remedy still outstanding, and an appellate decision could reshape the analysis.
  • How single-sex provision interacts with the 2026 EHRC Services Code after For Women Scotland is still working through, and has not been tested against specific facilities configurations.
  • Whether a lesser measure was genuinely available is decided on the facts of each case. No decision supplies a general test, and none is likely to.

Where this bites

The question in front of you is not “what is the rule” but “can I produce the assessment showing who this affects, what less restrictive option I considered, and why I chose as I did”.

Read the detail

“We are required to provide gender-neutral toilets. / We are not allowed to have them any more.”

Neither, and there is a real requirement underneath both that neither claim states. For a new non-domestic building in England, or a material change of use, Requirement T1 of the Building Regulations requires reasonable provision of separate male and female single-sex toilets; universal self-contained toilets are permitted in addition, and are lawful as the only provision only where space precludes the separate ones. That is a building-standards requirement, not the Equality Act question, and the two are routinely answered as one.

What is actually the case

Both versions of the claim are the same error, and it is an easy one. Everything to do with washrooms arrives as a single practical question — what do we put on the doors — so the three things that actually govern it get answered as one. The equality-law argument is the loudest of the three, so it tends to supply the answer to all of them.

The building standards come first, and they are the part most often missed. Requirement T1 of the Building Regulations, which Approved Document T supports, provides that toilet accommodation in buildings other than dwellings must consist of reasonable provision for male and female single-sex toilets, or — where space precludes that — universal toilets, and may consist of universal toilets in addition to single-sex ones. So there is a sex-based requirement in play, and it comes from building standards rather than from the Equality Act. It applies in England, to new buildings and material changes of use, and it excludes schools, early-years and custodial cellular accommodation. It sits on top of the workplace welfare regulations rather than replacing them. Read against that, the first claim is wrong — nothing obliges you to provide universal toilets — and so is the second, because they are expressly permitted alongside single-sex provision.

None of that is UK-wide, which is the trap underneath the trap. Approved Document T is England only. Scotland has no Part T equivalent: its Building Standards Technical Handbook at 3.12.1 says separate male and female sanitary accommodation is usually provided, proportionate to the building’s use, with unisex accommodation permitted only where each facility is a single-occupancy space for one person at a time, securable from inside. Wales has none either — its Approved Document G covers sanitary conveniences but imposes no single-sex mandate, so separate-sex provision in a Welsh workplace runs off regulation 20 of the 1992 Regulations, which applies across Great Britain. Northern Ireland has its own building control regime. So “do we have to” has a different answer in Manchester, Glasgow and Cardiff, and a different answer again for a workplace than for a building open to the public.

The Equality Act question is separate, and there the position genuinely is permissive: the exceptions that allow a service or facility to be restricted by sex are ones a provider may use, not ones it must discharge. Two questions, two sources, two different answers — which is exactly why collapsing them costs so much. An organisation that lets the equality-law argument settle its signage can end up non-compliant with a building standard nobody in the room was arguing about, and an organisation that reads Part T as settling the equality-law question has imported an obligation from the wrong statute. There is a third distinction underneath both: facilities for staff and facilities for the public sit under different rules, and an answer that is right for one can be wrong for the other in the same building. The order that holds up is to establish which regime governs and whether this is a new build or a material change of use, provide what the standards require, and treat any further restriction as a specific, proportionate, recorded exception. A restriction adopted because it is administratively simpler, because it feels safer, or because somebody is applying pressure is the least defensible option available.

What is genuinely unsettled

  • How the 2026 Services Code applies to particular premises configurations has not been widely tested, and the interaction between the staff and services regimes in shared buildings is a recurring practical gap.
  • Requirement T1 is England-only and reaches new buildings and material changes of use. What counts as space precluding separate provision, and how the requirement interacts with an existing building being altered rather than changed in use, is a question for a building-control professional rather than a policy.

Where this bites

Before you change a sign, establish whether this is a new build or a material change of use in England — because if it is, Requirement T1 has already decided part of the answer, and it decided it before the equality-law argument started.

Read the detail

Misconceptions

Belief

“Gender-critical belief is protected, so anything said in support of it is protected too.”

The belief is protected. How it is expressed is a separate question, decided on proportionality. Forstater established the protection; Higgs set out how a court approaches manifestation and proportionate disciplinary action; in Mackereth a health assessor who refused to use service users’ pronouns lost.

What is actually the case

The confusion is built into how the protection was reported. Forstater was a significant win and was covered as one, and the natural reading of “this belief is protected” is that acting on it is protected as well. For most protected characteristics that inference is roughly safe, because they are not things a person expresses in conduct towards colleagues. Belief is different: it is the one characteristic whose exercise routinely consists of saying things to other people.

So the law splits the question. Holding the belief is protected, and an employer that treats holding it as misconduct is on the wrong side of the line. Manifesting it is assessed on its own facts: what was said or done, to whom, in what role, with what effect, and whether the employer’s response was a proportionate answer to that conduct rather than a penalty for the belief behind it. Higgs is the Court of Appeal working through exactly that structure. Mackereth is the other end of it — refusing to use service users’ pronouns in a role that required addressing them was conduct, and the claim failed.

The cases do not all run one way, and an employer reading only the ones above would take the wrong lesson. In Adams v Edinburgh Rape Crisis Centre a tribunal found what it characterised as a heresy hunt and upheld direct discrimination and harassment against the employer. In Meade v Westminster City Council the claimant succeeded on direct discrimination, harassment and victimisation, with exemplary damages awarded over the conduct of the proceedings. An employer that treats the holding of the belief as misconduct, or that investigates the person rather than the conduct, is not in a grey area.

For an employer the practical consequence is that the question is never “is this belief protected”, because the answer is yes and it does not help you. The question is whether the specific conduct in front of you, in the role that person holds, can be addressed proportionately — and whether your response would look like a response to conduct or a response to the belief. Those look very different in a judgment, and the difference is usually visible in how the employer recorded its reasoning at the time.

What is genuinely unsettled

  • Where manifestation crosses into conduct an employer may act on is decided case by case. Higgs gives the framework, not a rule, and each new decision is on its own facts.

Where this bites

Address conduct, name the conduct, and record why your response was proportionate to it. A response that reads as a penalty for the belief is the one that fails.

Read the detail

“Protecting one side here means discriminating against the other.”

The framing is the error, not the answer to it. These are not two versions of the same claim waiting to be ranked: a belief is a view protected for whoever holds it, and being trans is a fact about a person rather than a proposition they argue for. Stop trying to rank them and what is left is a proportionality question, which has answers.

What is actually the case

The zero-sum version arrives pre-packaged, and usually from whoever is asking you to decide. It also matches the shape of nearly every other contested question a manager handles: two parties, competing accounts, one of them right. Grievances work that way, disciplinaries work that way, and most disputes a manager has ever resolved have worked that way. Presented with this one in the same wrapper, concluding that the job is to decide who is right is not a failure of reasoning — it is pattern-matching to a pattern that has always held before.

Where it fails is in what the ranking would require you to do. To weigh a belief against another person’s characteristic, you would have to reach a view on whether the belief is correct — and no employer has standing to do that, no tribunal is asked to, and nothing in the Act invites it. That is why meetings run on this framing do not conclude: they are attempting a task the law never set, and the participants correctly sense that no available evidence would settle it. The three things being confused — a belief, a protected characteristic, and the separate legal status a certificate confers — do different jobs and sit in different provisions, and the topic hub on protected belief sets that architecture out properly.

The practical consequence is a change of brief, and it is a smaller brief than the one people think they have been handed. You are not adjudicating a disagreement about sex and gender. You are deciding whether a particular arrangement — a rota, a facility, a piece of conduct, a form of words — is a proportionate way of pursuing an aim you can state, given the rights each person actually holds. That test comes from Bank Mellat v HM Treasury (No. 2), and it asks four things of you rather than one: whether the aim is weighty enough to justify limiting anybody’s rights at all, whether the measure you picked actually serves it, whether something less intrusive would have done the job, and whether what you gained is worth what it costs the person carrying it. Failing any one of the four fails the whole test — and every one of them is answerable on your own facts, which is what makes it a task rather than an argument. There is a usable tell in the room: if the discussion is about who is right, it is the wrong discussion and will not finish; if it is about which arrangement is defensible, it is the right one and it will.

What is genuinely unsettled

  • Whether a belief in gender identity attracts protection on the same footing has generally been accepted at tribunal level but not settled higher up, so the two sides of this are not the mirror image of each other that the zero-sum framing assumes.

Where this bites

If your meeting is trying to establish who is right, it will not conclude. Change the question to which arrangement you can defend, and it will.

Read the detail

Misconceptions

Conflict and challenge

“This is a safeguarding risk, so safeguarding has to come first.”

Safeguarding is a duty, not a trump card, and what it actually requires is a proportionate, evidence-led response to a specific risk — never a blanket assumption of risk attached to a group. Treating it as an override misreads it as badly as refusing to engage with it at all.

What is actually the case

The word carries real weight, and it should. Safeguarding obligations are serious, they are owed to children and adults at risk, and no responsible organisation wants to be the one that was warned and did nothing. It is also, for that reason, the most effective thing anybody can say in a meeting: it ends the discussion, because nobody wants to be recorded arguing against safeguarding.

What it requires is more specific than the way it is usually invoked. Three duties arrive together whenever a policy touches a service people use in person: the privacy of the trans person, the dignity of everyone using the service, and the safeguarding obligations the organisation already carried. They are not three competing interests to be resolved in favour of the loudest. A policy treating safeguarding as automatically overriding the other two has misunderstood it, because safeguarding is a proportionate response to a specific, evidenced risk rather than a blanket assumption. A policy treating privacy as absolute and declining to engage with safeguarding at all has made exactly the same error in the other direction.

The practical test is whether there is conduct to respond to. A safeguarding concern about what a named person has done is a safeguarding matter, and it is handled through the safeguarding process on its own evidence, as it would be for anybody. A concern about a category of person is not a safeguarding concern; it is a generalisation, and acting on it neither protects anyone nor survives scrutiny. The regulator’s own direction is to manage behaviour rather than police identity, and there is a practical reason as well as a legal one: no UK document reliably evidences sex at a service point, appearance evidences nothing, and challenging someone at a reception desk creates a fresh dignity and data-protection exposure of its own.

What is genuinely unsettled

  • How the 2026 Services Code applies to specific service configurations has not been widely tested, and safeguarding practice in this area is still being worked through by regulators rather than settled.

Where this bites

Ask what this person is alleged to have done. If there is an answer, run the safeguarding process on it. If there is not, what you have is a generalisation, and acting on it creates a second risk rather than removing the first.

Read the detail

“A colleague says our policy breaches their rights, so the policy must be wrong.”

A complaint is a signal that something needs answering, not a ruling that you were wrong. The first job is not to draft a reply — it is to route the correspondence correctly and protect the evidence trail behind the decision being complained about.

What is actually the case

Both reflexes are understandable and both are wrong. One is to defend immediately, because the challenge feels like an attack on the organisation. The other is to reverse immediately, because reversing makes the discomfort stop. The defensive reply is the one that reads worst later in a tribunal bundle or a board review; the instant reversal is the one that tells every future complainant that pressure is what works, and leaves the people affected by the reversal with a decision nobody reasoned.

Before either, the correspondence has to be routed, because a single letter is frequently more than one thing. A statement that a policy is discriminatory is a policy challenge. A solicitor threatening proceedings is a legal threat. A request for the personal data you hold is a subject access request under the UK GDPR, running on its own statutory timetable of one month rather than on yours. A grievance is a grievance under your own procedure. Each has a different process, a different audience and a different risk profile, and getting the routing right at the outset is the first decision to make — recorded, so you can show it was made deliberately rather than on the hoof.

The most valuable thing in the first forty-eight hours is not a response. It is retrieving the evidence trail behind the decision: what was asked, what was weighed, which options were considered, why this one was proportionate. That is far harder to reconstruct under pressure than to retrieve from where it was already sitting — and the complaint is the moment that separates organisations that built the trail beforehand from those that did not. Acknowledge receipt, say nothing substantive, then run the process: triage, assess proportionality, take advice, reply.

What is genuinely unsettled

Nothing in this one turns on an unsettled question.

Where this bites

Acknowledge within a day, say nothing substantive, and spend the first forty-eight hours finding the record rather than writing the reply.

Read the detail

Misconceptions

What the survey shows

“Everyone else has this sorted — we are the ones behind.”

Of 136 UK organisations surveyed for Beyond Compliance, 50.7% have no formal policy on trans or nonbinary inclusion at all. You are not behind the sector; the sector has largely not done this. Participation was voluntary, so that figure is most likely an optimistic floor.

50.7% of 136 UK organisations surveyed for Beyond Compliance (fieldwork November 2025 to January 2026) have no formal policy on trans or nonbinary inclusion at all. See the evidence

What is actually the case

The belief is not carelessness, it is a sampling problem, and it has a specific mechanism. The organisations you hear from on this subject are the ones with something to say about it — the conference speaker, the award entry, the campaign in June, the peer who mentions their new policy. Those are disproportionately the organisations furthest along. Nobody stands up at a sector event to describe the policy they have not written. So visibility reads as prevalence, and the most advanced fraction of the field becomes, in the mind of everyone watching, the normal state of it.

The measured picture is different. Of 136 UK organisations across the public, private and third sectors, 50.7% had no formal policy of any kind. That is not a gap between the leaders and the stragglers; it is the middle of the distribution. It also sits alongside a set of findings that describe the same absence from other angles: 41.0% with no formal process for a name or gender-marker change, and 30.6% estimating that fewer than one in ten of their staff would feel comfortable disclosing their identity.

What follows from that is not comfort, it is a change of task. If you believed you were catching up, the implied job was to find the established approach and adopt it. There is no established approach to adopt, no sector consensus to follow, and no template that has already been tested against a challenge. You are not behind a queue — you are making a decision, earlier than most, and the absence of anybody to copy is the actual condition of the work.

What is genuinely unsettled

  • Survey limits apply and cut one way. Participation was voluntary, so organisations willing to complete a survey on trans inclusion are plausibly more engaged than those that were not — which means these figures are most likely an optimistic floor rather than a worst case. Responses are self-reported, and fieldwork ran November 2025 to January 2026, before the 2026 Services Code came into force.

Where this bites

Stop looking for the approach everyone else has adopted. It does not exist, and looking for it is how organisations spend a year not deciding.

Read the detail

“Our general EDI policy already covers this.”

It may well name it. The commoner failure is not an absent policy but a present one with nothing underneath it: of 136 UK organisations surveyed for Beyond Compliance, 36.4% have policies and give managers no guidance on how to apply them.

36.4% of 136 UK organisations surveyed for Beyond Compliance have policies but give managers no guidance on how to apply them. See the evidence

What is actually the case

This is a reasonable position and often a deliberate one. A general policy genuinely does cover a great deal, a proliferation of single-issue policies is its own governance problem, and in policy terms “covered” properly means named. One document, one owner, one review cycle is a defensible design choice rather than an oversight.

The difficulty is that coverage in a document and capability in a corridor are different things, and the survey separates them. 36.4% of organisations have a policy but give managers no guidance on applying it. On training, 25.0% provide none on this topic at all and 33.8% fold it into general EDI content. So the modal organisation here is not one that failed to mention trans inclusion. It is one that mentioned it, and stopped — and that is precisely the Paper Shield: a policy with no manager guidance, no systems capability and no measurable outcome behind it. Its fragility is invisible until it is tested, which is the point at which the policy is examined for what happened when it met a real case.

The useful test is not whether the words appear. It is whether a line manager, told on a Tuesday morning that a team member is transitioning, knows what to do that day: what to change in which system, what to say to the team and with whose consent, what not to ask, what to record and where. If the answer to that is “they would read the policy”, read the policy yourself and check whether it answers any of it.

What is genuinely unsettled

  • Survey limits apply and cut one way. Participation was voluntary, so organisations willing to complete a survey on trans inclusion are plausibly more engaged than those that were not — which means these figures are most likely an optimistic floor rather than a worst case. Responses are self-reported, and fieldwork ran November 2025 to January 2026, before the 2026 Services Code came into force.

Where this bites

Test the policy against a Tuesday morning, not against a checklist. If it does not tell a manager what to do that day, it is a statement of intent.

Read the detail

“Someone senior owns this already.”

More often than not, nobody does. Of 136 UK organisations surveyed for Beyond Compliance, 53.7% have no one accountable at senior level for this area and only 6.1% tie inclusion outcomes to any executive measure — so even where there is a name, accountability is rarely attached to anything.

53.7% of 136 UK organisations surveyed for Beyond Compliance have no one at senior level accountable for this area; 70.9% have no named lead at all. See the evidence

What is actually the case

It feels owned, and that feeling has a cause. Somebody chaired the working group, somebody signed the policy off, somebody speaks at the network event. Each of those is a real act by a real person, and collectively they produce a strong impression that the area has a home. What they have in common is that they are episodes. Signing is mistaken for owning, because signing is the visible part.

The survey finds 53.7% of organisations with nobody accountable at senior level, and 70.9% with no named lead of any kind. Only 6.1% connect inclusion outcomes to executive KPIs or accountability measures, which means that even in the organisations that can name someone, the name usually carries no consequence. The effect is structural rather than personal. The person closest to a risk — a manager, a service lead, a data-protection officer — generally sees it first but lacks the authority to act, and the route upwards is undefined because nobody has been named to receive it. So the risk is noted, perhaps discussed, and left. Not through negligence: there is simply no threshold that says this goes higher, and nobody whose job it is to catch it.

The test is deliberately awkward. Name the person. Then say what happens to them if this goes wrong, and where it appears in their objectives. If the name takes a moment to produce, or the consequence is nothing, what exists is an interest rather than an owner — and an unowned risk has no escalation path, so it matures quietly until it is a crisis that could have been handled earlier for a fraction of the cost.

What is genuinely unsettled

  • Survey limits apply and cut one way. Participation was voluntary, so organisations willing to complete a survey on trans inclusion are plausibly more engaged than those that were not — which means these figures are most likely an optimistic floor rather than a worst case. Responses are self-reported, and fieldwork ran November 2025 to January 2026, before the 2026 Services Code came into force.

Where this bites

Name the person and name the consequence. If either takes more than a moment, the next risk in this area will sit exactly where the last one did.

Read the detail

“Our managers would know what to do if it came up.”

On their employers’ own assessment, 13.2% of 136 UK organisations surveyed for Beyond Compliance say their managers feel well-equipped and confident. That figure is the organisation rating itself, which is the direction of error you would expect.

13.2% of 136 UK organisations surveyed for Beyond Compliance say their managers feel well-equipped and confident; 30.9% report some awareness but a lack of practical guidance, and 22.8% say managers are not equipped at all. See the evidence

What is actually the case

Two things make this belief very stable. The first is that managers say they would be fine, because almost nobody volunteers that they would be out of their depth in a conversation about a colleague — the admission costs something and the situation is hypothetical. The second is rarity. For most managers this conversation happens once in several years, so the absence of an incident reads as evidence of competence when it is mostly evidence of infrequency. A capability you have never had to use looks identical to one you have.

The organisations themselves do not report confidence. 13.2% say their managers feel well-equipped; 30.9% report some awareness but no practical guidance; 22.8% say managers are not equipped at all. The direction of that error matters: this is the employer’s assessment of its own managers, which is the least likely party to understate the problem. And it lines up with what the case law shows, which is that most claims do not begin with a policy. They begin with a manager who did not know what to do, a concern that was not logged, or a decision nobody wrote down.

The first conversation is also the moment with the least margin. Protection begins when somebody says they intend to transition, so everything that follows — who was told, what changed in which system, what was asked that should not have been — is already inside the protected period, and it is happening in a corridor rather than in a governance forum. That is the hour of exposure, and it is the hour furthest from anybody senior.

What is genuinely unsettled

  • Survey limits apply and cut one way. Participation was voluntary, so organisations willing to complete a survey on trans inclusion are plausibly more engaged than those that were not — which means these figures are most likely an optimistic floor rather than a worst case. Responses are self-reported, and fieldwork ran November 2025 to January 2026, before the 2026 Services Code came into force.

Where this bites

Ask a manager what they would do in the first hour, not whether they would cope. The gap between those two answers is the exposure.

Read the detail