Malicious entities from foreign nations have begun using ChatGPT along with various AI models to engage in a range of cyber operations, according to a new OpenAI report.
According to the findings, users connected to China and Russia have been leveraging OpenAI’s technology, alongside models like China’s Deepseek, for phishing schemes and covert impact operations.
“There’s a growing trend where threat actors utilize a mix of AI models to pursue their objectives,” OpenAI noted.
A specific group of accounts linked to the Chinese government exhibited behaviors typical of intelligence operations, using AI to craft phishing materials in different languages while also creating malware and other tools.
They explored the possibility of employing DeepSeek for automation. In one instance, they analyzed online content to compile a list of potential email targets, tailoring their messages for maximum appeal.
OpenAI took action by banning the account, but they mentioned it couldn’t definitively identify whether the automation with other AI models had been used in the end.
Another set of accounts, based in Russia, crafted prompts to develop news-style videos, complete with scripts, SEO-friendly descriptions, hashtags, and translations using ChatGPT.
This seems to be part of a larger Russian influence campaign previously flagged by OpenAI, which involved posting AI-generated content on various platforms and social media. The latest posts criticized the actions of France and the U.S. in Africa, while praising Russia. Even though these accounts have been banned, they generated content that was largely critical of Ukraine and its allies, though such efforts failed to gain significant attention.
OpenAI also reported having banned several accounts suspected of belonging to the Chinese government. These accounts attempted to utilize ChatGPT for crafting proposals aimed at extensive surveillance, including tracking social media and movement.
“While these actions seem to stem from individuals rather than larger organizations, they offer a rare glimpse into how authoritarian regimes may exploit AI,” the company stated.
