Verify a decision

{"actor":"judge","decided_at":"2026-08-25T22:40:31.087Z","decision":"review_revise","investigation_id":"c1e79c32-018b-450b-855c-d8d387fea579","new_score":0,"page_slug":"dprk-crypto-theft-h1-2026-trm-labs-blockaid-report","prev_score":0,"reason":"Of 36 checked claims, 25 were confirmed, 6 partially supported, 4 disputed, and 1 unverifiable, putting the disputed-or-unverifiable share at about 13.9% -- within the range where a page is revised but not delisted. The issues that were found are concentrated in exactly the area this page depends on most: precise attribution to named research reports. The page's headline $643M/two-thirds figure is cited to a TRM Labs article that actually reports a different, earlier figure ($577M/76%, for a shorter year-to-date window), while the real source of the $643M number is a separate TRM Labs report the page never cites (claim_findings[0], claim_findings[3]). The page also overstates TRM Labs' and Elliptic's role in attributing the Drift Protocol attack to a named threat-actor subgroup with 'medium confidence' -- that specific, graded attribution reportedly came from Drift itself, SEAL 911, and Mandiant, with TRM Labs and Elliptic only corroborating a general North Korea link (claim_findings[8]). A $75 million figure for funds frozen by Arbitrum after the KelpDAO hack appears three separate times on the page, while independent outlets consistently report $71 million (claim_findings[17], claim_findings[23], claim_findings[34]). Finally, the page's summary presents 55% and 76% as if they were two contemporaneous, alternate methodology-based estimates for the same H1 2026 period, when the 76% figure is an earlier, narrower snapshot that the same firm's own later report superseded -- a more misleading framing than simply citing a stale number, since it implies an unresolved disagreement between sources where none exists. None of this undermines the page's core narrative -- the Drift, KelpDAO, and Humanity Protocol incidents, the OFAC sanctions action, and the general scale of DPRK crypto theft are all well supported -- but the sourcing precision failures are numerous enough and concentrated enough in load-bearing claims to warrant a moderate penalty and revision rather than approval.","score_delta":-13,"sequence_num":3,"submission_content_hash":null,"submission_id":null,"submission_kind":null,"submission_valence":null,"v":1}