Existing antiarrhythmic drugs to treat atrial fibrillation (AF) have incomplete efficacy, contraindications and adverse effects, including proarrhythmia. AP30663, an inhibitor of the K2 channel, has demonstrated AF efficacy in animals; however, its efficacy in humans with AF is unknown. Here we conducted a phase 2 trial in which patients with a current episode of AF lasting for 7 days or less were randomized to receive an intravenous infusion of 3 or 5 mg kg AP30663 or placebo. The trial was prematurely discontinued because of slow enrollment during the coronavirus disease 2019 pandemic. The primary endpoint of the trial was cardioversion from AF to sinus rhythm within 90 min from the start of the infusion, analyzed with Bayesian statistics. Among 59 patients randomized and included in the efficacy analyses, the primary endpoint occurred in 42% (5 of 12), 55% (12 of 22) and 0% (0 of 25) of patients treated with 3 mg kg AP30663, 5 mg kg AP30663 or placebo, respectively. Both doses demonstrated more than 99.9% probability of superiority over placebo, surpassing the prespecified 95% threshold. The mean time to cardioversion, a secondary endpoint, was 47 (s.d. = 23) and 41 (s.d. = 24) minutes for 3 mg kg and 5 mg kg AP30663, respectively. AP30663 caused a transient increase in the QTcF interval, with a maximum mean effect of 37.7 ms for the 5 mg kg dose. For both dose groups, no ventricular arrhythmias occurred and adverse event rates were comparable to the placebo group. AP30663 demonstrated AF cardioversion efficacy in patients with recent-onset AF episodes. K2 channel inhibition may be an attractive mechanism for rhythm control of AF that should be studied further in randomized trials. ClinicalTrials.gov registration: NCT04571385 .