Verify a decision

{"actor":"judge","decided_at":"2026-08-26T00:04:37.008Z","decision":"review_revise","investigation_id":"bba44da0-a72e-4e24-8e86-51fcdfad5505","new_score":0,"page_slug":"global-pig-butchering-enforcement-cluster-276-arrests-m-seizures-2026","prev_score":5,"reason":"Recounting the reviewer's 31 claim findings by hand (16 confirmed, 5 partially supported, 6 disputed, 4 unverifiable) and applying the disputed-plus-unverifiable formula yields 10/31 = 32.3%, just over the 30% line into the significant-issues band; the reviewer's own summary block undercounted total claims as 30 but the underlying array and math check out. The disputed findings are not minor date slips: the page's centerpiece section attributes a $701.96 million crypto restraint, 503 seized domains, a Telegram channel takedown, and device counts to the April 29 arrests, when the review shows independent reporting ties those same figures to a separate DOJ announcement six days earlier, and the April 29 release itself discloses no dollar figure at all (claim_findings[5], [9], [10], [11]). A phrase describing a Thai arrest as a distinct 'additional fugitive' obscures that the person is one of the four defendants named two sentences later (claim_findings[7]). Separately, the page asserts a February 26 aggregate figure 'encompasses' three later, and in one case larger, recoveries — which is not possible on the page's own numbers (claim_findings[28]). These are reinforced by two high-priority coverage gaps flagging the same attribution and chronology problems. Weighing against denial: every individual enforcement action, defendant, and charge is independently confirmed against primary sourcing, no charged individual is described as convicted, and the reviewer's confidence (0.72) is moderate rather than low, arguing for revision rather than the more severe outcomes.","score_delta":-20,"sequence_num":3,"submission_content_hash":null,"submission_id":null,"submission_kind":null,"submission_valence":null,"v":1}