Net Good IndexSubmit a correction

Harm Ledger · provisional · Economics & Productivity

OpenAI disrupts a Cambodia-based ChatGPT scam-compound operation

On 4 August 2026 OpenAI reported banning a coordinated ChatGPT account network, likely operating around Poipet, Cambodia, that used the models for romance, investment, gambling, and law-enforcement impersonation scams and for compound administration, after a WhatsApp lead.

4 Aug 2026Tier 2 Significant HarmMethodology 0.1

Current score

−0.19

3 base · Significant Harm (tier 2 of 5, 3 pts)
× 0.4500 attribution · Material acceleration
× 0.5000 evidence · External expert evaluation
× 0.7000 realization · Independently validated or deployed
× 0.4000 durability
Event-level product before credit split: 0.19

Documented AI-enabled scam operations with unverified loss totals are significant but not population-scale harm (tier 2). Human compounds used ChatGPT as a production tool (0.45). First-party disruption report plus industry journalism (0.50). Realization is reported ongoing scams later banned (0.70).

What happened

OpenAI says operators generated fake personas, translated outreach on WhatsApp and Telegram, forged document images, and kept internal records of worker debts and immigration status. Some chats referenced detention and forced labor consistent with public reporting on Southeast Asian scam compounds. OpenAI could not independently verify claimed victim losses; it described evidence of hundreds of targets and individual claims of thousands of dollars. This is AI-enabled crime, not an autonomous fraud agent.

Model attribution

−0.19

GPT

Generated personas, translations, promotional copy, forged-document images, and internal compound administration text.

OpenAI attributes the banned account cluster to ChatGPT use; human operators ran the scams.

Attribution 0.4500 · Credit share 100% · OpenAI

Claims

  • OpenAI banned a coordinated Cambodia-linked ChatGPT account network used for romance, investment, gambling, and impersonation scams.

    outcome · supported

  • OpenAI independently verified a precise total of stolen funds.

    significance · disputed

Sources

primary sources

independent sources

Secondary domains: Safety

Revision history

  • 13 Sep 2026 · 0.00 → 0.19

    Imported events under methodology 0.1.

OpenAI disrupts a Cambodia-based ChatGPT scam-compound operation · NetGoodIndex