BACKGROUND: The potential of HIV self-testing (HIVST) to cause harm is a concern hindering widespread implementation. The aim of this paper is to understand the relationship between HIVST and harm in SELPHI (An HIV Self-testing Public Health Intervention), the largest randomised trial of HIVST in a high-income country to date. METHODS: 10 111 cis and trans men who have sex with men (MSM) recruited online (geolocation social/sexual networking apps, social media), aged 16+, reporting previous anal intercourse and resident in England or Wales were first randomised 60/40 to baseline HIVST (baseline testing, BT) or not (no baseline testing, nBT) (randomisation A). BT participants reporting negative baseline test, sexual risk at 3 months and interest in further HIVST were randomised to three-monthly HIVST (repeat testing, RT) or not (no repeat testing, nRT) (randomisation B). All received an exit survey collecting data on harms (to relationships, well-being, false results or being pressured/persuaded to test). Nine participants reporting harm were interviewed in-depth about their experiences in an exploratory substudy; qualitative data were analysed narratively. RESULTS: Baseline: predominantly cis MSM, 90% white, 88% gay, 47% university educated and 7% current/former pre-exposure prophylaxis (PrEP) users. Final survey response rate was: nBT=26% (1056/4062), BT=45% (1674/3741), nRT=41% (471/1147), RT=50% (581/1161).Harms were rare and reported by 4% (n=138/3691) in exit surveys, with an additional two false positive results captured in other study surveys. 1% reported harm to relationships and to well-being in BT, nRT and RT combined. In all arms combined, being pressured or persuaded to test was reported by 1% (n=54/3678) and false positive results in 0.7% (n=34/4665).Qualitative analysis revealed harms arose from the kit itself (technological harms), the intervention (intervention harms) or from the social context of the participant (socially emergent harms). Intervention and socially emergent harms did not reduce HIVST acceptability, whereas technological harms did. DISCUSSION: HIVST harms were rare but strategies to link individuals experiencing harms with psychosocial support should be considered for HIVST scale-up. TRIAL REGISTRATION NUMBER: ISRCTN20312003.