OpenAI banned a cluster of accounts linked to Spamouflage, a Chinese state-linked influence operation, after catching it using ChatGPT for research and website debugging.
According to OpenAI, the accounts belonged to Spamouflage and used its models to research activity on social media platforms, generate posts, and debug code on a website that had not previously been linked to the operation. Investigating that AI usage is what led OpenAI to the website in the first place. OpenAI removed the accounts after identifying the activity as part of the operation.
The interesting part here isn't the propaganda - it's the plumbing. Spamouflage didn't use AI to make its messaging more persuasive; it used AI as a research assistant and an unpaid junior developer, the same way any ordinary team might. That's a duller, more realistic picture of AI misuse than the deepfake headlines suggest - less about generating convincing fake video and more about quietly speeding up the unglamorous grunt work behind an influence campaign.
Banning accounts is the easy part. The next unreported website won't announce itself so conveniently.