OpenAI links China-based accounts to US data-centre influence push
OpenAI says it banned accounts it assesses as China-linked after they used ChatGPT to draft posts and cartoons aimed at US arguments over AI data centres. The company says the operation, labelled Data Center Bandwagon, tried to amplify existing concerns about electricity prices, grid capacity and local impacts, but did not gain meaningful traction. The episode matters beyond the United States because AI infrastructure has become a strategic bottleneck: compute capacity, power supply and public consent now sit inside the same geopolitical contest. For Belgian and EU readers, the signal is not that all opposition to data centres is foreign-made. It is that real local concerns about energy, land and water can become useful material for influence operations. The EU's AI Act and Brussels-based AI Office are moving into enforcement at the same time as governments face pressure to expand AI capacity without weakening democratic debate.
Verified by Validiris·📚 7 sources·🧠 AI-checked·🇧🇪 Belgian: MediumWhy you can trust this
About this story
OpenAI (US artificial-intelligence company founded in 2015, maker of ChatGPT) is the entity making the attribution claim. ChatGPT (OpenAI's conversational AI service launched in 2022) was allegedly used to draft influence content. Data Center Bandwagon (OpenAI's label for the suspected operation) refers to content about AI data-centre costs and grid strain. Ben Nimmo (OpenAI Intelligence and Investigations principal investigator and former influence-operations researcher) briefed reporters on the case. X (social platform formerly known as Twitter) and Facebook (Meta-owned social network) were named as places where content appeared. The EU AI Office (European Commission office in Brussels created to implement the AI Act) oversees general-purpose AI obligations at EU level. The Artificial Intelligence Act (EU Regulation 2024/1689, adopted in 2024) is the bloc's main AI rulebook. AI data centres (high-density computing facilities for training and running AI models) are the infrastructure at the centre of the dispute.
How to read this story
The history
OpenAI and other technology firms have reported earlier misuse of generative AI for influence operations, but OpenAI says this case is notable because it targeted the data-centre debate. The broader pattern predates generative AI: state-linked influence campaigns have long latched onto divisive domestic issues rather than inventing them from scratch. The EU's response has also evolved: Regulation (EU) 2024/1689 entered into force in 2024, while the Brussels-based AI Office was set up to supervise general-purpose AI obligations. Energy concerns are equally old, but AI has made them sharper by concentrating demand in specific grid locations.
The geopolitics
Compute infrastructure is now part of strategic competition. The United States, China and the EU all treat AI capacity as economically and militarily significant, while data centres depend on chips, electricity, water, land and public permission. Influence activity around infrastructure debates can therefore serve a geopolitical goal even when the visible message looks local and economic.
Why now
The trigger is OpenAI's June 2026 disclosure that it had disrupted accounts it links to China-based actors using ChatGPT for data-centre messaging. The timing also follows a surge in political attention to AI infrastructure, energy demand and local opposition in the United States and Europe.
What to watch
Watch whether OpenAI releases fuller technical indicators, whether other model providers report similar data-centre influence attempts, and whether EU AI Office guidance treats misuse reporting as part of systemic-risk governance. In Belgium, future large data-centre proposals would test how local consultation handles suspicious online amplification.
Regional impact
The EU level is most directly affected because the European Commission's AI Office supervises general-purpose AI obligations and will help shape platform and model-provider expectations. Belgium is affected mainly as host state for EU institutions and as a small, grid-constrained economy where data-centre siting would be politically sensitive. Flanders, Wallonia and Brussels would face different practical questions if major AI infrastructure projects emerged: land availability, grid capacity, permitting rules and heat-reuse potential differ by region.
Local impact
The most local Belgian relevance is for municipalities and industrial zones that may face future data-centre siting decisions, especially around grid access and land use. No Belgian project is named in the reported OpenAI case, so the practical lesson is procedural: local consultations need transparent participants and clear evidence on energy costs.
International angle
The story sits inside the US-China technology contest, but Europe is not a bystander. EU institutions in Brussels are writing and enforcing rules for general-purpose AI, while European governments also want domestic AI capacity. The same data-centre arguments over cost, sovereignty and sustainability can travel across borders quickly.
What this means for you
For Belgian and EU readers, nothing changes immediately in law or energy bills. The practical takeaway is to scrutinise provenance in infrastructure debates: who funds campaigns, whether accounts are authentic, and whether claims about power prices or grid strain are backed by regulators, grid operators or permitting documents.
What happens next
OpenAI is likely to keep publishing disruption findings as influence operators test generative tools. EU regulators could watch whether model providers document misuse patterns before AI Act enforcement tightens. Belgian and EU authorities should expect future infrastructure debates to include both genuine local concerns and attempts to amplify them through inauthentic accounts.
Potential consequences
If the pattern spreads, data-centre permitting debates in Europe could become harder to read: authentic objections, corporate lobbying and covert amplification may appear in the same online spaces. That could push governments toward stricter disclosure rules, stronger platform cooperation or more centralised infrastructure planning. It could also backfire if officials dismiss legitimate community concerns as foreign manipulation without evidence.
Opposing perspectives
- OpenAI Intelligence and Investigations team
OpenAI's framing is that the campaign shows foreign operators testing how generative AI can scale content around existing US political flashpoints. The company says the operation was small and ineffective, but still important because it reveals the narratives and tactics being explored by China-linked influence actors.
- Community and environmental data-centre opponents
Local opponents would argue that concerns over electricity bills, grid capacity, water use and land impacts do not become illegitimate because foreign actors try to exploit them. The strongest version of this view is that transparent permitting, cost allocation and environmental scrutiny are the antidote to both corporate overreach and covert amplification.
- EU technology regulators
The EU regulatory frame is that general-purpose AI providers need governance systems that address systemic risks without suppressing lawful debate. EUR-Lex records the AI Act's aim as protecting fundamental rights, democracy and environmental interests while supporting innovation, so the regulatory problem is both misuse prevention and public trust.
Timeline
- 2024-07-12·EUR-Lex published Regulation (EU) 2024/1689, the Artificial Intelligence Act, in the Official Journal.
- Late 2025 to early 2026·OpenAI says the suspected Data Center Bandwagon activity took place during this period.
- 2026-06-10·OpenAI briefed reporters and accounts of the China-linked data-centre influence campaign were published.
- 2026-06-11·The story entered international news feeds as a data-centre and geopolitical influence issue.
Glossary
- General-purpose AI
- AI models that can be adapted to many tasks, such as text generation, coding or image analysis, rather than one narrow use.
- EU AI Office
- A European Commission office in Brussels responsible for helping implement the AI Act, especially rules for general-purpose AI models.
- AI data centre
- A high-density computing facility designed to train or run AI models, often requiring large power supplies and specialised cooling.
- Influence operation
- A coordinated attempt to shape public debate, often by hiding who is behind the messaging.
Related to this story
Live connections from the Belgium Impulse ecosystem — not recommendations.
This briefing was prepared with AI assistance and reviewed by a Belgium Impulse editor before publication. methodology.