OpenAI has taken action against a small network of ChatGPT accounts involved in an influence operation dubbed "VAGue Focus." These accounts used OpenAI's models to generate social media posts, analyze datasets, and translate communications for social engineering purposes. The activity was primarily conducted in Chinese during mainland Chinese business hours.
The operation involved generating messages that purported to originate from three geopolitically focused entities: "Focus Lens News," "BrightWave Media Europe," and "Visionary Advisory Group" (VAG). The threat actors described these entities as fronts for intelligence collection and analysis.
Key Points
- OpenAI banned a network of ChatGPT accounts for using AI in social engineering, surveillance-themed research, and influence activities.
- The operation, named "VAGue Focus," used OpenAI's models to generate social media posts and translate messages from Chinese to English.
- Accounts prompted models in Chinese and were active during mainland Chinese business hours.
- The operation supported four main workstreams, including generating social media content for online personas and translating correspondence.
- Users also requested translations of messages designed to extract information from unknown interlocutors.
- The primary objective appeared to be intelligence collection by posing as professionals based in Europe or Turkey.
- The public-facing part of the operation showed little authentic engagement, with one X account's follower count attributed to a compromised and repurposed account.
Context
According to OpenAI's June 2025 report, the network used ChatGPT to create social media posts and biographies for online personas distributed by X accounts posing as journalists and geopolitical analysts. The accounts also polished and translated correspondence addressed to a US Senator. Additionally, they asked basic questions about computer network attack and exploitation tools, suggesting a low level of expertise. Users regularly requested translations of messages from Chinese to English that appeared designed to extract information.
Why It Matters
This case highlights the ongoing challenge of preventing the misuse of AI models for influence operations and social engineering. It demonstrates how AI can be leveraged to create deceptive online personas and facilitate intelligence collection, prompting developers to implement robust safety measures and monitoring.
What To Do
- Note the types of malicious activities, such as social engineering and influence operations, that AI models can be used to support.
- Review the details of the "VAGue Focus" operation to understand the tactics employed, such as generating social media posts and translating deceptive communications.
- Watch for indicators of compromised social media accounts, as described in the case study, when assessing online engagement.
- Consider the implications of AI being used for intelligence collection and the need for ongoing vigilance in model deployment.