Data Center Bandwagon: an attempt to look like grassroots criticism
OpenAI reports fabricated American personas spreading criticism of data-center costs. It found no evidence of meaningful spread beyond the operation itself.
The short version
- In June 2026, OpenAI described likely China-origin accounts using invented American identities and AI-generated comments and images.
- The campaign targeted data-center electricity costs; the accompanying overview found no meaningful spread beyond its own activity.
Why it matters. The reported deception was about who was speaking: coordinated accounts tried to pass as ordinary participants in a local debate.
Real news, invented neighbors
A legitimate regional newspaper story about electricity-market prices became material for an influence campaign. OpenAI says the operators requested comics and comments linking rising costs to AI and data centers, then posted AI-assisted content on X.
OpenAI assessed the operators as likely connected to a private Chinese technology company serving provincial government clients. Original case report, “Actor” and “Behavior”
The ordinary-neighbor disguise
Work reports submitted for polishing described the machinery behind the personas: start with everyday interests, have accounts interact to make engagement appear organic, and keep backups ready. The reports also discussed Facebook pages and advertising. They reveal a planned distribution strategy, not proof that every proposed ad ran.
A campaign is not the same as a result
The decisive limit: OpenAI’s June overview found no evidence of meaningful spread beyond the operation’s own activity. The record describes an attempted intervention, not demonstrated success in changing public opinion.
See the examples: the case report’s Platform operations section includes posts and edited images from the investigation.
A dated record
- June 2026
The case appears in OpenAI's threat reporting.
OpenAI · source ↓ - June 10, 2026
The overview explains the limits of the observed impact.
OpenAI · source ↓
The supporting record
OpenAI describes accounts likely originating in China using fabricated US personas and AI-assisted comments and images about data-center costs.
OpenAI assesses the operators as likely linked to a private technology company serving provincial-level Chinese government clients.
OpenAI's June overview reports no evidence of meaningful spread beyond the operation's own activity.
OpenAI says work reports described building everyday personas, arranging interactions between accounts and maintaining backup accounts.
What we do not yet know
- Independent reproduction of the operator attribution.
- Any connection to the separately reported August X network.
- Evidence of a downstream policy or construction outcome.
Original documents, with context
Sources you can inspect
1. Platform investigation · OpenAI
Data Center Bandwagon: US-targeted influence activity
OpenAI describes activity it investigated on its services. Attribution language remains qualified; this is not independent reproduction of its investigation.
Publication: June 2026 · Checked: 2026-09-07
Source record: openai-bandwagon
Location in source: Actor; Behavior: newspaper-based content and work reports; Platform operations: examples of posts and edited images2. Platform disclosure · OpenAI
PRC-linked influence operations are targeting AI debates in the US
The overview explicitly limits evidence of spread beyond the operation. It is related reporting from the same publisher, not an independent second investigation.
Publication: 2026-06-10 · Checked: 2026-09-07
Source record: openai-june-overview
Location in source: Campaign descriptions and observed impact
Corrections & additional evidence · support@aipublicopinion.com
