OpenAI Disrupts Operation High Five Influence Campaign

OpenAI has banned multiple ChatGPT accounts associated with Comm&Sense Inc, a commercial marketing company in the Philippines, for violating policies against using AI to generate bulk political influence content. The operation, dubbed "High Five," used ChatGPT to analyze social media trends, generate partisan comments in English and Taglish, and create promotional materials for the covert operation.

Operation High Five: Actor and Methodology

Comm&Sense Inc utilized ChatGPT to automate several stages of a political influence campaign targeting audiences in the Philippines. The actor used the model to perform content analysis on social media posts regarding political events—specifically those involving President Bongbong Marcos—to determine appropriate themes for replies.

Once themes were established, the actor used ChatGPT to generate bulk volumes of short comments, typically under ten words, to be posted across social media platforms. Additionally, the model was used to generate PR pitches and statistical analyses to market the covert influence operation to current and existing clients.

Platform Distribution and Execution

The operation deployed content across TikTok and Facebook using the following strategies:

  • TikTok: The operation managed five dedicated TikTok channels that began posting in mid-February 2025. These channels posted identical videos with varying captions, and dozens of TikTok accounts were used to post AI-generated comments in reply to those videos.
  • Facebook: AI-generated comments were posted in reply to news reports from mainstream media outlets. OpenAI noted that there was no indication that these mainstream outlets were connected to the operation.

Content Characteristics and Messaging

The generated content was brief and partisan, primarily focusing on praising President Marcos and his initiatives or criticizing Vice-President Sara Duterte. Some comments specifically referred to Vice-President Duterte as "Princess Fiona," likely referencing the Shrek movie series. While the majority of the content was political, a minority of the completions included commercial marketing materials, suggesting the firm was operating on behalf of multiple clients.

Impact and Effectiveness

Despite posting thousands of comments across TikTok and Facebook, the operation had limited organic reach. OpenAI's investigation found that none of the identified comments received more than single-digit likes, shares, or replies, with most receiving none at all.

Using the Brookings Institution's Breakout Scale for measuring influence operations, OpenAI categorized this activity as Category 2. This classification indicates that while the activity occurred across multiple platforms, there was little evidence that real people widely shared or engaged with the content.

Following the initial ban of the accounts, OpenAI observed several attempts by the threat actor to return to the models based on technical and behavioral indicators.

Sources