Overview

• OpenAI reports that adversaries often plan with ChatGPT and execute with other AI models such as DeepSeek or Claude, which limits investigators’ visibility into full operations.
• Accounts that appeared tied to Chinese government entities were banned after seeking proposals for large-scale monitoring, including a tool to analyze Uyghur travel patterns and police records.
• OpenAI also removed Chinese-language accounts exploring phishing automation on DeepSeek and banned Russian-speaking criminal groups that used ChatGPT to refine malware and generate prompts for covert influence videos.
• Investigators conclude the activity reflects incremental efficiency gains to established tactics, and say they have disrupted more than 40 malicious networks since February 2024.
• The report notes simple evasion tricks to mask AI-generated writing and minimal traction for identified influence campaigns, as users increasingly turn to ChatGPT to help spot scams.