OpenAI bans China-linked ChatGPT accounts running US tariff and data-center influence campaigns

OpenAI disclosed it banned a set of China-linked accounts that used ChatGPT to draft content for social-media influence campaigns aimed at US public debates over tariffs and AI data-center policy, characterizing it as a 'likely' Chinese influence operation. The disclosure is part of OpenAI's ongoing threat-reporting cadence around state-aligned misuse of its tools.

The mechanism: rather than the models being jailbroken, the accounts used ChatGPT as a content-generation engine to scale propaganda — drafting posts and talking points designed to shape opinion on politically sensitive economic topics. OpenAI's response was account-level enforcement plus public attribution.

This lands the same week Google sued a China-based group for Gemini-powered fraud and the US government acted against Anthropic's models — a striking convergence of frontier labs being pulled into national-security and influence-operation policing. It reinforces a narrative of AI providers as front-line actors in geopolitical information conflicts.

OpenAI also expanded its Trusted Access for Cyber program with a new GPT-5.4 Cyber model, signaling a vetted-access approach to dual-use security capabilities — a contrast to Anthropic's Mythos, which just ran into government intervention. Watch how attribution standards hold up and whether enforcement meaningfully deters state-scale abuse.