OpenAI says it disrupted Cambodia-based scam ring using ChatGPT

OpenAI says it disrupted a Cambodia-based scam operation that used ChatGPT to produce multilingual scripts and messaging for investment, romance, gambling and impersonation schemes, according to the company’s disclosure on July 28, 2026. The company did not identify individual victims or name local operators but described removing accounts and content connected to the network in coordination with platform enforcement tools, saying the disruption prevented further abuse.OpenAI’s blog post

The announcement follows months of reporting that AI tools have been repurposed by organised fraud networks across Southeast Asia. OpenAI framed the intervention as part of a broader effort to police malicious uses of its models, while external investigators and rights groups warn the scale of Cambodia’s scam economy—and state responses—remain contested.Reuters reported in June that Amnesty had identified 86 scam sites in Cambodia, of which 24 had been targeted by authorities.

OpenAI's claim: accounts removed, multilingual scripts blocked

OpenAI’s post says the operation used ChatGPT to generate “conversation scripts, investment pitches, and impersonation messages” across multiple languages and to scale social-engineering workflows. The company describes a combination of internal detection signals and manual review to identify and disable accounts tied to the ring, but offers limited technical detail about the detection heuristics or how attribution to a Cambodia-based network was established. Where OpenAI is the sole source, its wording is properly described as a company claim rather than independently verified fact.OpenAI’s blog post

Past reporting and government material, meanwhile, have already linked mainstream large language models to scam activity in Asia. A U.S. congressional research summary cites Reuters investigations that found ChatGPT among the most-used tools in fraud operations, underlining that the OpenAI disclosure is the latest episode in a documented trend of AI-assisted scams.U.S. congressional research citing Reuters.

Amnesty and Reuters: 86 sites, contested enforcement in Cambodia

Reuters’ June 8 reporting, which summarises Amnesty International’s findings, is the clearest independent context for OpenAI’s claim. That piece said Amnesty had identified 86 suspected scam sites and criticised state responses as “reactive and ineffective,” while Cambodian officials pointed to revocations of 25 casino licences, prosecutions of roughly 1,500 suspects from 19 countries, and the deportation or voluntary departure of tens of thousands of foreign workers.Reuters. The discrepancy between Amnesty’s critique and Cambodian authorities’ account highlights a governance gap that technology companies’ takedowns cannot by themselves fill.

Reporters and investigators also note a practical limit: platform-level disruption addresses the supply of automated content and accounts, but it does not dismantle the physical infrastructure—compounds, coercive recruitment networks and on-the-ground logistics—that power many large-scale fraud operations. Reuters’ longer investigations into the region show those facilities remain active even after police actions.Reuters reporting.

OpenAI’s disclosure therefore satisfies part of the “what happened” ledger: it documents a company intervention against abuse of its product. It does not, however, settle the broader questions about the operation’s size, the identity of operators, or the degree to which national authorities are complicit or effective—questions Amnesty and Reuters continue to press.Reuters.

External critics also caution against over-relying on platform removals as evidence of success. Amnesty’s findings suggest many sites evade enforcement; independent journalists have documented large compounds and complex logistics that take more than account suspensions to dismantle.Reuters.

OpenAI’s move underscores a larger industry problem: mainstream LLMs have become tools in criminal economies, and companies face pressure both to harden models and to share more operational detail when they intervene. The next concrete measure to watch is whether Cambodian authorities corroborate the link OpenAI describes and whether follow-on prosecutions or cross-border investigations—cited by Reuters as ongoing—produce public evidence tying specific facilities or operators to the AI-enabled campaigns.OpenAI’s blog post