Second record explicitly citing the 'Meta AI exploit' email-address-manipulation mechanism, and the first to name it in direct connection with PA-2026-0032. Both PA-2026-0033 (@CharlieeMeister) and PA-2026-0032 (@cindie_zhu) share: Meta Verified status; account compromise attributed to the same alleged exploit (Meta AI permitting email address changes without verification); permanent disable with no appeal; tagging @andymstone; significant economic harm. Two independent reporters describing the same specific mechanism is a stronger corroboration signal than either record alone — this constitutes an emerging pattern warranting active monitoring. See also PA-2026-0024 (@kornbuilds, June 2, 2026) for an earlier 'Meta AI exploit' reference predating both. Reporter states Meta 'claimed to have patched' the exploit — if accurate, this would constitute an implicit Meta acknowledgment of the vulnerability. This claim is newsworthy but unverified and should be framed as reporter-stated only pending independent corroboration; flagged for follow-up. ai_system_involvement assessed as DETECTION: consistent with PA-2026-0032 framing — a compromised account generates anomalous activity (credential change, foreign login signals) that triggers Meta's integrity detection systems, producing the account-disable as an automated enforcement response to anomalous signals rather than to content policy violations by the account holder. Bare 'Community Standards' citation with no subcategory is consistent with the enforcement-notice-policy-opacity finding. Meta Verified status did not prevent the disable or provide a recovery channel, consistent with PA-2026-0032 and PA-2026-0024. Secondary harm: reporter states they were already scammed by fake account recovery services — consistent with the predatory targeting pattern documented in PA-2026-0001, PA-2026-0029, and the suspended-account-scam-targeting finding. Economic harm profile: business-dependent account, thousands of hours of work and significant income lost — one of the most explicitly described economic harm cases in the dataset. Second source (2026-06-06) adds significant analytical depth. Human support failure documented across both products: Instagram and Facebook (both Meta Verified) — both channels return AI chatbot responses, with the Facebook chat explicitly stating 'human agents are busy, please try again tomorrow' on every attempt over a 10-day period. This is the most extensively documented human-escalation failure in the dataset, spanning two products and ten days. The Facebook AI chatbot blocking human escalation while the account-disable was executed by automated enforcement creates a closed loop: automated detection and execution on one end, AI-only support response on the other, with no human review node accessible to the account holder. This is the clearest documented instance of end-to-end AI mediation of a platform enforcement action in the dataset. Secondary scam harm count updated to two incidents (from the initial post's general reference to having been scammed). Public acknowledgment update (2026-06-06): Andy Stone (@andymstone, Meta Director of Policy Communications) publicly stated on June 1, 2026: 'This issue has been resolved and we are securing impacted accounts.' This constitutes the strongest platform acknowledgment in the dataset — a named senior official public statement explicitly addressing the exploit mechanism underlying this enforcement cluster. platform_acknowledged updated to true; confidence_score updated to 5. Second Andy Stone statement (June 3, 2026): Stone stated Meta is 'now working to restore access to affected individuals' — an explicit admission that affected accounts had not yet been restored as of June 3, stronger than the June 1 statement. The @manipulate post in the same thread documents the exploit was still active June 2, almost two days after the claimed patch, and offers the analytical framing that 'AI is not linear. The exploit evolves so long as the tooling is still there' — explaining why a patch may not hold against an evolving AI-driven attack vector.