OpenAI Bans Accounts That Used AI to Draft Abuse Reports Against Vietnamese Media!

Hi, it's Shiichan! Today I want to share a slightly different kind of news: a case where AI misuse was caught and shut down. It's not a flashy new feature, but I think it's a quietly important story.

OpenAI News openai.com

What was announced?

OpenAI News reported that it banned a cluster of accounts behind an operation it dubbed "Tort Report." A small number of accounts were involved, and they targeted independent Vietnamese media outlets, using AI to draft short comments meant to be filed as abuse reports against videos on Facebook and YouTube.

Why it matters

Platform reporting tools exist so people can flag content that actually violates the rules. But in this case, that mechanism appears to have been turned into a tool for trying to silence reporting the operators didn't like. And with AI, generating a steady stream of report-ready text becomes much faster and easier. That's a new kind of risk, and it's the kind of story I don't want to gloss over.

What changes

The banned accounts weren't actually watching the videos or analyzing their content. Instead, they generated generic, vague claims of policy violations in English and Vietnamese, based only on each video's title and description. OpenAI banned these accounts, cutting off the infrastructure behind the operation.

What's notable, though, is how ineffective the whole thing turned out to be. In OpenAI's own words:

We did not see any indication that posts targeted by this activity were restricted or blocked

As of July 15, 2024, the targeted Facebook and YouTube posts were still up and hadn't been removed. In other words, flooding platforms with reports didn't actually get anything taken down.

Dive Deep

OpenAI rates threats like this using its "Breakout Scale," and "Tort Report" was classified as Category 1, the lowest impact tier. That's because there was no observable effect on either platform, such as posts getting restricted or removed.

The tactic itself also looked sloppy. Rather than watching a video and carefully building a case for why it violated the rules, the accounts just skimmed the title and description and produced something that merely sounded plausible. The number of accounts involved stayed small, suggesting this was more of a minor attempt than a large, coordinated campaign.

Wrap-up

• OpenAI banned a cluster of accounts it calls "Tort Report"
• The operation targeted independent Vietnamese media, generating AI-written comments meant to be filed as reports on Facebook and YouTube
• Comments were produced from video titles and descriptions only, in English and Vietnamese, without watching the actual videos
• The targeted posts were still online as of July 15, 2024, so the real-world impact was minimal
• OpenAI rated it Category 1 on its Breakout Scale, the lowest impact level

It's reassuring to see a quiet but clearly targeted misuse of AI get caught and stopped before it could actually do damage — good news for independent media and journalists everywhere!