Chinese bot farms and AI data centers: the influence operation on X
- X neutralized approximately 200,000 inauthentic accounts linked to China.
- 200 specific accounts manipulated the debate on energy costs of AI data centers.
- The strategy used real grievances (electricity bills) combined with AI-generated content.
- The operation follows a pattern similar to an OpenAI takedown that occurred in June 2026.
The expansion of artificial intelligence infrastructure is no longer just a technological or energy supply challenge, but has become a battleground for geopolitical influence operations. The recent dismantling of a network of coordinated accounts on X reveals how foreign actors are seeking to sabotage AI industrialization in the United States, not through cyberattacks, but by manipulating social discontent.
The weapon of 'real discontent': how Chinese bots rode the price of electricity
The strategy adopted by the Chinese bot farm was not based on the invention of total falsehoods, but on the amplification of pre-existing tensions. The accounts used what can be defined as 'real discontent': the increase in domestic electricity bills and the pressure on local energy grids.
According to what was reported by X, the operators rode the narrative that AI data centers were driving the increase in energy costs for citizens. To give credibility to these claims, the campaign integrated references to legitimate reports, such as an article from the 'Southern Maryland Chronicle' regarding the increase in capacity costs of PJM Interconnection. PJM, which serves 13 states and Washington, D.C., had confirmed in a June 2026 market update that data centers are the primary driver of new energy demand growth.
Strategic Analysis: This technique is extremely dangerous for businesses because it does not fight facts, but distorts the perception of them. Instead of denying the existence of data centers, the operation transformed a technical grid management problem into a class conflict, positioning AI infrastructure developers as 'villains' who enrich themselves at the expense of families.
200,000 accounts for 200 messages: the asymmetry between bot-farm volume and target precision
The numbers emerging from the report by X's Global Government Affairs team highlight a surprising operational asymmetry between the critical mass of the bot-farm and the targeted influence action.
| Metric | Value | Operational Function |
|---|---|---|
| Total accounts neutralized | ~200,000 | Creation of background noise, support infrastructure, simulation of mass. |
| AI/Energy targeted accounts | ~200 | Production of specific content to manipulate public debate. |
| Volume/Precision Ratio | 1000:1 | Use of a vast network to protect and amplify a core of surgical messages. |
This structure suggests that most of the accounts served to create an ecosystem of 'synthetic authenticity', providing likes, reposts, and interactions to make the 200 active accounts appear as influential voices supported by a real community, when in reality they were part of a single coordinated apparatus.
From PJM reports to AI-generated comics: the anatomy of hybrid disinformation
The operation was configured as a hybrid disinformation campaign, capable of mixing technical data, emotional narratives, and synthetic content. The content creation process followed a precise flow:
- Anchoring Phase: Use of real data (e.g., June 2026 PJM reports) to establish an indisputable base of truth.
- Distortion Phase: Direct and simplified link between data center energy demand and the immediate increase in domestic bills.
- Emotionalization Phase: Production of visual material via AI. Cartoons and comic strips were created depicting data center operators as greedy figures, contrasted with families in economic difficulty.
The effectiveness of this approach lies in its ability to bypass critical thinking: a user who reads a real piece of data (the cost of energy) is more likely to accept the resulting satirical image, perceiving the entire narrative as coherent.
Thomas Jackson and the 'synthetic citizens': the aesthetics of counterfeit American identity
To make the influence credible, the bot farm created 'synthetic citizens', profiles that perfectly imitated the identity of ordinary American citizens. The chosen names were deliberately generic and reassuring, designed not to raise suspicion and to integrate into the digital social fabric of the USA.
Among the identified profiles were names such as Thomas Jackson, Patricia Robinson, Sandra Thompson, and Kenneth Flores. These accounts did not act as automatic bots repeating the same message, but 'argued', simulating discussions and personal opinions to manipulate the debate on energy policies and AI.
Business Analysis: The creation of these synthetic identities represents an evolution of social engineering. It is no longer about phishing to steal passwords, but about 'identity laundering' to inject political opinions into strategic markets, potentially influencing local zoning decisions or energy concessions for data centers.
Influence vs Cyberattack: why manipulating public debate is more effective than hacking the network
The case highlighted by X raises a fundamental point about the nature of modern threats. As emphasized in the sources, the accounts in question did not hack an electrical grid nor did they breach the systems of a utility.
- Cyberattack (Hard Power): Aims to disrupt service, cause blackouts, or steal data. It is an overt act, often attributable, and generates an immediate and coordinated security response.
- Influence Operation (Soft Power): Aims to change public opinion. In this case, the goal was to push Americans to oppose the construction of AI data centers.
'It was an influence campaign, not a cyberattack, and that distinction is the whole story.'
The strategic advantage of manipulation is invisibility: while a cyberattack activates defense protocols, an influence campaign exploits freedoms of expression and internal social divisions so that the population itself blocks the technological development of its own country.
From OpenAI to X: the chronology of cleanup operations against Asian interference
X's operation is not an isolated event, but fits into a pattern of monitoring and removal activities of coordinated influence networks.
- June 2026: OpenAI carries out a takedown of similar influence networks. Some of the AI-generated images used in the data center campaign had already appeared in an OpenAI report.
- August 28, 2026: X's Global Government Affairs team identifies and neutralizes the bot farm of approximately 200,000 accounts.
- August 30, 2026: Dissemination of details on the operation and analysis of manipulative content regarding energy and AI.
Data centers, energy sovereignty, and AI Act: the risk of social sabotage for critical infrastructures in the EU
Although the described operation targeted the United States, the implications for the European Union and Italian companies are direct and immediate. Europe is facing the same transition toward AI and the same energy tensions, making the continent fertile ground for identical tactics.
Implications for businesses and the regulatory framework
For entrepreneurs investing in data centers or AI solutions in Italy and the EU, the risk is not only technical (cybersecurity), but reputational and social. The manipulation of public debate can lead to:
- Administrative blocks: Increase in local opposition (NIMBY) fueled by bot farms, which can slow down construction permits.
- Regulatory pressure: Push toward more restrictive energy regulations based on distorted perceptions of actual consumption.
- NIS2 and AI Act risk: While the AI Act focuses on the safety and ethics of models, and the NIS2 directive on the resilience of critical infrastructures, neither regulation fully addresses 'social sabotage' via external disinformation.
Future scenarios and monitoring indicators
It is likely that these campaigns will shift toward other critical sectors of the ecological transition. For businesses, it is fundamental to monitor the following indicators:
- Anomalous Sentiment Indicators: A sudden spike in local opposition to an infrastructure project, characterized by standardized language and the use of AI-generated visual content (Indicator: Local social media analysis).
- Correlation between Technical Reports and Memes: The appearance of satirical content that oversimplifies technical reports from regulatory bodies within 48 hours of their publication (Indicator: Temporal monitoring of content).
- Evolution of the AI Act: The eventual introduction of specific clauses on the transparency of synthetic content used in political pressure campaigns (Indicator: EU legislative updates 2027).
In conclusion, European technological sovereignty is not defended only with hardware and software, but with the ability to protect public debate from synthetic manipulation.
FAQ
How many accounts were involved in X's operation?
X neutralized approximately 200,000 accounts suspected of being inauthentic and linked to China, of which about 200 were specifically dedicated to manipulating the debate on AI data centers.
What was the main objective of the disinformation campaign?
The objective was to push American citizens to oppose the construction of data centers for AI, claiming that these facilities were causing the increase in domestic electricity bills and risking the stability of the energy grid.
How did the bots make their claims credible?
They used a hybrid strategy: they cited real reports (such as those from PJM Interconnection) and created visual content via AI, such as comics and cartoons, to emotionalize the message and depict AI developers as greedy.
Sources: Startupfortune, Nextbigfuture, Techspot, Yahoo · by glacom.news AI
Scrivila qui: Susanna, l assistente AI di glacom, ti risponde via email con un approfondimento gratuito.
Nessuna consulenza personalizzata (finanziaria, legale o medica): solo informazione e fonti. Email usata solo per rispondere.
.png)
