Tort Report: Abusive reporting activity

OpenAI's latest threat report reads less like a policy update and more like a dispatch from an ongoing skirmish. Buried in the details: a cluster of accounts that had been leaning on ChatGPT to churn out abusive reports and complaints aimed at Vietnamese public figures and platforms. Not malware, not disinformation memes — just an assembly line for grievance, dressed up to look legitimate.

The mechanics are almost mundane, which is what makes them unsettling. Instead of one person laboriously typing out a complaint, a bad actor can ask a language model to generate dozens, varying tone and phrasing enough to dodge spam filters or human reviewers. Scale up, and you get a coordinated harassment campaign that looks, on the surface, like a groundswell of authentic complaints rather than a handful of operators with a chatbot.

OpenAI says it caught and banned the accounts responsible. Good. But the fact that this tactic showed up on their radar at all says something about where AI misuse is headed next. Forget flashy deepfake videos for a second — the more mundane threat is AI as a force multiplier for petty, targeted harassment, the kind that platforms already struggle to moderate even without machine-generated volume behind it.

Vietnam isn't named as a coincidence. Reports like this tend to surface in places where public figures already face intense, sometimes state-adjacent, pressure campaigns, and where platforms may have thinner moderation resources than in the US or EU. That combination — motivated actors, real-world targets, and a moderation gap — is exactly the sort of environment where an LLM-as-complaint-generator scheme finds oxygen.