OpenAI busts 'Sneer Review' op that used ChatGPT to flood TikTok with fake Taiwan outrage

OpenAI disrupted a covert Chinese influence operation that was leaning heavily on ChatGPT to manufacture outrage, dubbing it “Sneer Review.” The network’s most distinctive work was generating dozens of critical comments and a long-form article slamming “Reversed Front,” a Taiwanese video and board game about resisting the Chinese Communist Party. But the bad actors didn’t stop at gaming culture wars. They also used the models to bulk-generate social media posts in English, Chinese, and even Urdu across TikTok, X, and Reddit, creating a false sense of grassroots engagement.

One particularly ugly facet involved targeting Pakistani activist Mahrang Baloch, a vocal critic of Chinese investments in Balochistan. A TikTok account and Facebook Page linked to the network posted a video falsely accusing Baloch of appearing in pornography, then flooded the post with hundreds of AI-generated comments to simulate widespread condemnation. The operation also chimed in on the closure of USAID, generating both praise and criticism to muddy the waters. Internally, the actors were using ChatGPT for performance reviews, with one user outright claiming in a prompt to work for the Chinese Propaganda Department—a detail OpenAI says it cannot independently verify.

What’s striking is just how amateurish this was. On Facebook, the two Pages posing as news outlets had zero followers. A Reddit post scraped just 44 upvotes while others were removed by moderators. The TikTok videos did amass a combined 25,000 likes and tweets hit around 10,000 views, but OpenAI is skeptical. Because the network itself was generating the majority of the comments on its own posts, the company warns that all engagement figures should be treated with heavy caution.

The operation was clearly in its infancy when OpenAI pulled the plug. The network’s sloppy tradecraft—like using a Korean screen name to post an Urdu comment, or crypto-branded X accounts with cartoon avatars—made the artificial engagement easy to spot once you looked. Using its own IO impact scale, OpenAI places this at the low end of Category 3 only if you take the engagement numbers at face value. If those likes and views turn out to be mostly AI-generated noise too, the real-world impact drops even further. It’s a reminder that much of this activity is less a sophisticated psy-op and more a clumsy attempt to simulate a crowd from an empty room.