OpenAI Flags China AI Influence Ops

OpenAI has disrupted what it describes as two distinct influence operations originating from China, leveraging its AI models to manipulate public discourse in the United States. The company announced the findings, detailed in a report published June 10, 2026, highlighting attempts to interfere with critical debates surrounding artificial general intelligence and broader tech policies.

The first operation, dubbed the "Data Center Bandwagon" campaign, generated social media comments and images alleging that AI data center expansion was driving up electricity costs for ordinary families. This campaign aimed to exploit existing public anxieties regarding energy prices and the local environmental impact of large-scale data center development.

A second cluster, named the "Tech and Tariffs" campaign, focused on criticizing US tariffs as tools for technological dominance. Notably, prompts for this campaign specified the exclusion of Chinese leader Xi Jinping and the inclusion of President Trump, suggesting a targeted political framing. This cluster was also linked to a network of inauthentic social media accounts that falsely claimed ChatGPT user data had been compromised, an allegation OpenAI refuted.

While OpenAI found no evidence that these operations significantly swayed public opinion, their significance lies in the targeting of AI infrastructure itself. This represents an effort by state-backed actors to probe and potentially undermine the foundations of US technological leadership and its democratic AI ecosystem. Such AI influence operations aim to exploit societal divisions and erode trust.

OpenAI's disclosure aims to bolster the collective ability of industry, governments, and civil society to identify and counter foreign threat actors seeking to manipulate democratic debates and advance authoritarian agendas through AI. The company's mission includes ensuring AI benefits all of humanity, which necessitates vigilance against its misuse for surveillance, censorship, and control.