The Prevent Paradox

How Britain’s counter-radicalisation duty drifted into a false-positive system for autistic and SEND children — and why the state is now on dangerous ground.

There is a moment in some school meetings when the air changes.

Nothing dramatic. No raised voices. No obvious crisis. Just a sentence that lands oddly. A child answers too literally. Or talks too long. Or fixates on the wrong detail. Or says something clumsily political because they are trying to be precise, not provocative.

Someone in the room feels a flicker of discomfort — not because anything violent has been said, but because it doesn’t fit the expected script. And in a system built around fear of missing the one-in-a-million event, not fitting starts to feel like evidence.

Someone remembers their Prevent training. Someone else remembers “duty of care”. The safeguarding lens quietly rotates into something else: risk. A form is opened. Not because anyone is convinced. Because nobody wants to be the adult who didn’t escalate.

From that point on, the child is no longer being understood primarily as a child who needs support. They are being processed as a potential security concern.

That is the Prevent paradox in miniature: a programme sold as early help ends up functioning as institutional anxiety made official.

Prevent’s purpose, on paper, is straightforward. It asks public bodies to identify vulnerability to radicalisation early and to intervene through support, not punishment. Channel exists as the mechanism for that support. In abstract form, this sounds like safeguarding. Nobody serious disputes the idea that genuine risk should be addressed.

The problem is not the existence of Prevent. The problem is what happens when you graft a counter-terrorism duty onto everyday institutions that are already overstretched, under-specialised, and structurally terrified of being blamed.

Because Prevent does not operate in a calm environment of measured judgement. It operates in classrooms, SEND reviews, pastoral systems, and local authority pipelines where uncertainty is routine and defensive decision-making is rational.

Ask one blunt question and the logic becomes obvious: who gets punished for being wrong?

If a teacher refers a child “just in case” and the case goes nowhere, the teacher looks cautious. If they don’t refer and something later happens — however tenuously connected — the teacher looks negligent. That imbalance is not a moral failure. It is an incentive. And incentives shape behaviour far more reliably than guidance documents. Add a second, unavoidable reality: terrorism is rare. Autism and SEND traits are not.

Any system that trains large numbers of non-specialists to “spot early indicators” of an extremely rare outcome will generate false positives unless thresholds are brutally high and evidence standards are unforgiving. If you make the system emotionally urgent but statistically vague, misclassification is guaranteed.

This is where autistic and SEND children get caught.

Many neurodivergent traits sit awkwardly with school norms: literal language, intense interests, rule-based thinking, blunt honesty, difficulty modulating tone, a tendency to explore ideas deeply without social padding. In a calm environment, these are recognised as support needs. In an anxious one, they are reinterpreted as “fixation”, “rigidity”, “obsession”, “isolation”.

The system doesn’t need prejudice to cause harm. It just needs fear plus poor calibration.

That miscalibration is reinforced by a SEND landscape that has been under strain for years. Neurodevelopmental assessments take too long. Specialist support is thin. EHCP processes are slow and exhausting. Schools are expected to cope with complexity they are not resourced for.

When intended pathways are blocked, pressure finds another outlet. Prevent becomes, quietly, a functional escalation route: a defined process with statutory weight, a way of triggering multi-agency attention when everything else feels stalled.

That alone should set off alarms. A counter-terror programme should never become a workaround for a broken disability support system. But systems do what their incentives encourage, not what their mission statements promise.

Defenders of Prevent often reach for the annual statistics at this point. Referral numbers. Channel cases. Closures. But the most important part of the story is what the data doesn’t show.

It does not consistently record disability or neurodivergence across referrals. It does not publish meaningful equality impact analysis. It does not track what happens to children who are referred and then closed — the distress, the stigma, the loss of trust, the signal sent to families that asking for help can backfire.

That absence is not a technical oversight. It is where the legal risk lives.

Public bodies are required to consider equality impacts. Not rhetorically. Not symbolically. Under the Public Sector Equality Duty, they must show that they have thought about how policies affect disabled people and adjusted accordingly.

Running a system that predictably captures disabled children, while failing to properly measure or mitigate the impact, is not a neutral position. It is a choice. And it is exactly why organisations such as Rights & Security International have begun challenging the Home Office on Prevent’s monitoring and impact assessment.

Courts do not ask whether harm was intended. They ask whether it was foreseeable — and what was done about it.

There is also a quieter, longer-term cost. Once institutions start treating vulnerability as suspicion, trust drains away. Families learn to be cautious. Children learn that curiosity, bluntness, or identity can trigger escalation. Staff learn that self-protection matters more than nuance.

Which brings us to the most sensitive part of the picture: identity and geopolitics.

In a polarised climate, institutions get jumpy. Jumpy institutions over-interpret. When cultural competence is thin and fear of under-reacting is high, identity itself can start to look like risk.

For neurodivergent children — who often speak plainly and without social buffering — the danger of misinterpretation increases. Talking about family history. Expressing interest in living in Israel. Discussing aliyah factually. Mentioning conscription as a legal reality rather than a political stance.

None of this is extremism. But a system trained to scan for “signals” and uncomfortable with ambiguity can misread it.

There is limited public data quantifying this. That is not a rebuttal. It is the problem. Under-measurement does not equal absence; it signals a failure of curiosity and care.

Now to accountability.

Prevent sits with the Home Office. Whatever its origins, the government of the day owns how it operates, what it measures, and how it responds to harm. Labour did not design Prevent, but Labour is now responsible for its consequences.

So far, the response has leaned heavily on process. Thresholds tightened. Oversight strengthened. Commissioners appointed. Guidance refreshed. All of this sounds reassuring. None of it fixes the underlying design failure.

You cannot solve an incentive problem with better paperwork.

As long as escalation is rewarded, as long as neurodivergence is treated as a vague vulnerability marker rather than a context requiring competence, and as long as equality impacts remain under-examined, the system will keep doing what it currently does — and ministers will keep inheriting the fallout.

At some point, “it’s meant to be safeguarding” stops being a defence and starts sounding hollow.

What a non-stupid system would do instead

Here is the uncomfortable truth: the fixes are not exotic. They are basic design hygiene. The reason they’re politically awkward is that they force government to admit the programme is currently producing predictable harm.

1. SEND-first triage If a child is known or suspected to be neurodivergent or have SEND needs, the default pathway must be support (SEND/pastoral safeguarding) — not security — unless there is specific, evidenced intent and capability.

   
   

2. Mandatory neurodiversity-competent review No Channel escalation for under-18s without a checkpoint review by someone trained to interpret neurodivergent presentation properly. Not “a bit of awareness training”. Competence.

   
   

3. Measure harms and publish them Collect equality-relevant data ethically, including disability markers and outcomes. Track false-positive proxies, the rate of closed cases, and wellbeing impacts. If you don’t measure harm, you are choosing ignorance.

   
   

4. Training reform: stop teaching ‘traits’ as indicators Stop turning characteristics into suspicion. Teach base rates, context, and decision hygiene: what evidence is required, what is merely awkwardness, what is SEND, and what is genuinely security-relevant.

   
   

5. Accountability loops, not referral dumping Local systems should be able to audit referrals, see outcomes, and learn calibration over time. A system that can’t learn will keep repeating superstition.

The paradox, restated

Prevent exists to stop harm before it happens. Used badly, it creates harm now — through mislabelling, stigma, self-censorship, and institutional mistrust — while offering the comforting illusion that something is being done.

A safeguarding duty that trains adults to suspect vulnerable children is not safeguarding. It is institutionalised anxiety with a government logo.

If the state genuinely wants to reduce risk, it needs to stop confusing escalation with wisdom. It needs to measure what it is doing to disabled children. And it needs to redesign the system so that thoughtfulness is not punished and fear is not rewarded.

Until then, Prevent will keep generating false positives — and the government will keep owning the consequences.

RW

---

Appendices

Appendix A — Glossary

• Prevent: UK counter-radicalisation programme within CONTEST.

• Channel: Multi-agency support process used for Prevent cases assessed as appropriate.

• SEND: Special Educational Needs and Disabilities.

• EHCP: Education, Health and Care Plan.

• PSED: Public Sector Equality Duty (Equality Act 2010, s149).

Appendix B — Evidence note

This piece separates:

• Published official material (government guidance/statistics).

• Credible reporting and legal advocacy (investigations, legal actions, rights organisations).

• Behavioural inference (incentive design and base-rate reasoning).

• Reported experience (flagged as under-measured where systematic data is not publicly available).

Appendix C — Systemic risks if no reform occurs

• Legal exposure under equality duties (failure to evidence due regard / monitoring).

• Continued misrouting of SEND cases into security frameworks.

• Chilling effect on expression and identity.

• Erosion of trust between families and schools.

• Political ownership of preventable harm by the government of the day.

Appendix D — FOI angles

• Prevent referral counts by age band and setting (primary/secondary/alternative provision).

• Outcomes: proportion closed, proportion to Channel, reasons for closure.

• Local training materials used (versions, suppliers, any content referencing neurodivergence).

• Equality impact assessments and monitoring practice.

• Any local audit reports, lessons learned, or governance minutes relating to Prevent referrals involving children.

Links & further reading

• Home Office statistics: Individuals referred to and supported through the Prevent Programme, April 2024 to March 2025 (published 6 November 2025).

• Prevent duty guidance (schools and other specified authorities).

• Shawcross: Independent Review of Prevent and government response.

• Rights & Security International: monitoring/impact concerns re autism and Prevent (analysis + documents).

• Legal pressure reporting: neurodivergent people over-reported; equality duty concerns.

• Children’s Commissioner: neurodevelopment waiting times (SEND context).

• Hansard: Parliamentary debate references on Prevent operation/thresholds/oversight.

• Office of the Independent Prevent Commissioner (oversight architecture).