What Active Bystander Training Is
A bystander is anyone present when conduct crosses a line who is neither the target nor the person doing it. An active bystander interrupts, checks on the person targeted, or reports it. In a survey of English NHS Trusts published in 2023, Robertson and Steele describe active bystander training as training that has been promoted to address sexual harassment, alongside other forms of poor behaviour, discrimination, and harassment. Acas, in its guidance on witnessing sexual harassment, says you could try to stop it if you feel it is safe to do so, should talk to the person who experienced it to see if they want your support, and calls it a good idea to make a note with dates, times and names. Each is something a person does, under pressure, with colleagues watching.
Provision is thin: of the 199 Trusts that responded, 35 provided active bystander training, most of it not specific to sexual misconduct and opt-in only. Knowing that you could interrupt is not the capability of interrupting.
The Legal Basis: Section 40A, and 30 October 2026
The duty is one sentence. Section 40A of the Equality Act 2010 reads: "An employer (A) must take reasonable steps to prevent sexual harassment of employees of A in the course of their employment." It was inserted by the Worker Protection (Amendment of Equality Act 2010) Act 2023, extends to England, Wales and Scotland, and has been in force since 26 October 2024. Under section 124A, a tribunal that finds sexual harassment may add up to 25% to compensation where section 40A was also breached.
On 30 October 2026 the bar rises. Section 20 of the Employment Rights Act 2025 inserts the word "all" before "reasonable steps". The section is not yet in force: the date comes from the Government's announcement, and commencement regulations are still awaited. The government's implementation timeline, updated 25 September 2026, lists under that date the "all reasonable steps" duty and "introducing an obligation on employers not to permit the harassment of their employees by third parties". Among the steps to consider, Acas's note on the change lists "sexual harassment training that is suited to your organisation, making use of specialist delivery where needed". More in our piece on the October 2026 bar and in harassment prevention training in London.
No statute names active bystander training; whether a programme is a reasonable step is for the organisation and its advisers to judge, and a bystander who has practised the interruption is a different piece of evidence from a completion certificate. FCA-regulated firms carry a second set of expectations on the same conduct, covered in our guide to the FCA non-financial misconduct rules.
Why a Policy Does Not Produce a Bystander
Darley and Latané's 1968 paper carries the mechanism in its title: Bystander intervention in emergencies: diffusion of responsibility. Fischer and colleagues' 2011 meta-analysis in Psychological Bulletin found an overall effect of g = -0.35: the presence of other people reduces the likelihood that any one of them helps. The effect was weaker where the situation was perceived as dangerous, a perpetrator was present, or the cost of intervening was physical; a remark in a meeting is not dangerous and the cost of interrupting it is social, although the person who made it is in the room and the others are usually colleagues rather than strangers, two conditions under which the meta-analysis also found the effect weaker. A policy puts no sentence in anyone's mouth for a moment that lasts about three seconds, which is why the duty needs behaviour rather than a document.
A module also feels as if it worked. In the randomised 2020 UCL study behind our method, 57 participants learned the same feedback skill through rehearsal with a role-play coach, slides or video, and rated their own confidence before the intervention, after it, and after a simulated conversation. Confidence rose in every group: slides 4.46, 5.00 and 5.39; video 4.69, 5.13 and 5.38; practice 4.27, 5.33 and 5.53 (Laumann (2020), UCL, unpublished research paper, Figure 4). Modality had no significant effect on confidence (p = .958), nor did the interaction of time and modality (p = .541). Observed delivery, rated blind by two independent assessors, differed: rehearsal came out ahead of slides (Cohen's d = 0.75) and of video (d = 1.08). The study tested one skill, its practice condition was a 20-minute online role-play coaching session, and it was not this programme; we do not present it as a measurement of it. It does suggest why a feedback form is a weak guide: self-rated confidence rose in every group, whatever the observed delivery.
Four Moments, Rehearsed as Scenes
Scenarios are written from the organisation's own material: a short diagnostic establishes where conduct crosses the line in this workplace, and the actors are briefed on those specifics. A typical bystander design would build around four moments, each a scene with one sentence the participant has to say, out loud, to an actor who responds as a real person would, followed by feedback and a second attempt. The lines below are illustrations, not a script.
The remark in the meeting. A comment about a colleague's body or private life lands and the agenda moves on. The move is one flat sentence in the next half-second: "Let's not do that. Back to the numbers." Rehearsal would work on speed and tone: ten seconds late, or delivered as a lecture, and the room moves the wrong way.
The senior person nobody challenges. The remark comes from whoever signs off the budget or the appraisal, and the sentence has to hold without becoming a confrontation or an apology: "I'd rather we didn't talk about her like that." The scene would also rehearse the fallback Acas describes, reporting it, for the day when saying it in the room is not safe.
The client who crosses the line. From 30 October 2026 employers must also not permit harassment by third parties, and the client in the meeting room is where that will be tested. The sentence would go to the client, in front of the colleague it was aimed at: "I'm going to stop you there. We don't do that here."
Afterwards, with the person it happened to. Acas's guidance is to ask whether they want your support; the conversation goes wrong when the bystander decides for them. The scene would rehearse the offer without the takeover: "I saw what happened. Do you want me to say something, or would you rather I didn't?" Then the note: what, when, and who else was there.
Format, Cohort and Measurement
Sidestream is a behaviour change consultancy based at 58 Malden Road, London NW5 3HG; most of this work is delivered in-house at client offices across London. A typical format is a half-day of four hours with 12 to 25 participants and one or two scripted scenarios; a participant can also bring a conversation of their own, the one still owed to a colleague after a moment they witnessed, and brief the actor themselves. Bystander rehearsal alongside disclosure response and manager handling of concerns is the one-day, multi-target format on our harassment prevention page. Work is priced per engagement, because the effort sits in the scenario design; why the other person in the scene is a professional actor is set out in role-play training with professional actors. The cohort that matters is the one present when a remark lands, more often a whole team than its managers.
Measurement sits at Kirkpatrick Level 3, observed behaviour back at work, rather than Level 1 satisfaction. The measures are those our harassment page names, observed bystander-intervention frequency and, where appropriate, downstream harassment-incident reporting; a typical design would read these through figures most HR systems already hold, such as the share of concerns that arrive from witnesses rather than only from the person targeted, and the time between an incident and its first report, baselined before delivery and re-read around ninety days afterwards. Setting up Level 3 is covered in how to measure behaviour change and the Kirkpatrick model.
Active Bystander Training in London: In Short
In England, Wales and Scotland, section 40A requires reasonable steps to prevent sexual harassment; from 30 October 2026 that means all reasonable steps, plus a duty not to permit harassment by third parties. A policy describes a bystander; it does not produce one. Active bystander training in London therefore has to be rehearsal: scenes with professional actors, one sentence each, attempted, coached and attempted again, in a half-day for 12 to 25 people, priced per engagement, and measured on observed behaviour rather than on how confident people felt on the way out. The wider programme, with six behavioural targets, is harassment prevention training in London; the same interruption aimed at bias is Target 2 of DEI training in London; what counts as harassment is in what is workplace harassment in the UK; raising things before they become incidents is speak-up culture training and how to build a speak-up culture. This page is general information, not legal advice.
Frequently Asked Questions
What is active bystander training?
Training for the people who witness conduct that crosses a line, so that they interrupt it, check on the person it was aimed at and report it, as Acas describes in its guidance on witnessing sexual harassment. Sidestream runs it as rehearsal with professional actors, so the interrupting sentence has been said out loud before it is needed.
Is active bystander training a legal requirement in the UK?
No statute names it. In England, Wales and Scotland, section 40A of the Equality Act 2010 requires employers to take reasonable steps to prevent sexual harassment of their employees, in force since 26 October 2024; from 30 October 2026 that becomes all reasonable steps, and employers must also not permit harassment by third parties. Whether a programme is a reasonable step is for the organisation and its advisers to judge.
Does it cover harassment by clients and other third parties?
Yes. From 30 October 2026 employers must not permit the harassment of their employees by third parties, and a typical design would make it one of the four moments: a professional actor in the client's chair, the participant stopping it in the room, then the second conversation with the colleague afterwards.
How do you know the training changed anything?
Not from a feedback form: in the randomised 2020 UCL study behind our method, self-rated confidence rose in every teaching condition, including the two whose observed delivery blind assessors rated lower. Measurement sits at Kirkpatrick Level 3 instead; a typical design would use figures the organisation already holds, such as the share of concerns raised by witnesses and the time to first report, baselined before delivery and re-read around ninety days afterwards.
Sources: Equality Act 2010, section 40A (employer duty to prevent sexual harassment of employees), legislation.gov.uk · Worker Protection (Amendment of Equality Act 2010) Act 2023, as enacted (extent: England, Wales and Scotland), legislation.gov.uk · Equality Act 2010, section 124A (compensation uplift of no more than 25%), legislation.gov.uk · Employment Rights Act 2025, section 20 (employers to take all reasonable steps to prevent sexual harassment; not in force at Royal Assent), legislation.gov.uk · Department for Business and Trade, Plan to Make Work Pay and Employment Rights Act: timeline update (updated 25 September 2026) · Acas, Sexual harassment: harassment law changes under the Employment Rights Act 2025 (26 August 2026) · Acas, Witnessing sexual harassment (25 March 2026) · Robertson, A. and Steele, S. (2023), A cross-sectional survey of English NHS Trusts on their uptake and provision of active bystander training including to address sexual harassment, JRSM Open 14(4), doi 10.1177/20542704231166619, PMC10164257 (responses collected by August 2022) · Fischer, P. et al. (2011), The bystander-effect: a meta-analytic review on bystander intervention in dangerous and non-dangerous emergencies, Psychological Bulletin 137(4), 517-537, PubMed · Darley, J. M. and Latané, B. (1968), Bystander intervention in emergencies: diffusion of responsibility, Journal of Personality and Social Psychology 8(4), 377-383, PubMed