X says it disrupted about 200,000 suspected Chinese inauthentic accounts, including roughly 200 that tried to push Americans against AI data centers by using a real grievance: higher power bills.
On Aug. 28, X's Global Government Affairs account said its Safety team had identified and disrupted a network of about 200,000 suspected Chinese inauthentic accounts. About 200 of them, X said, were posting in a way that could manipulate the American debate over AI data centers and energy policy. The accounts didn't hack a grid or breach a utility. They argued. It was an influence campaign, not a cyberattack, and that distinction is the whole story.
The bait was already there #
According to X's disclosure, the accounts posed as ordinary Americans. The names: Thomas Jackson, Patricia Robinson, Sandra Thompson, Kenneth Flores. They pushed the claim that AI data centers are driving up household electricity bills and straining local power grids. Some of the material leaned on legitimate reporting, including a Southern Maryland Chronicle article about PJM Interconnection's rising capacity costs. New data center demand is putting pressure on the grid, too. Then the accounts added the cheap stuff: AI-generated cartoons and comic strips showing data center developers as villains getting rich while families stared at higher power bills.
You don't need a bot farm to be angry about your electric bill. That is the uncomfortable part. PJM serves 13 states and Washington, D.C., and its own June 2026 market update said data centers are the primary driver of new demand growth because facilities can be built two to three times faster than much of the power generation needed to serve them. In March, the Southern Maryland Chronicle reported that PJM's 2024 auction sent capacity costs from about $2 billion to nearly $15 billion for the following year, before the next auction reached just over $16 billion under a cap requested by Pennsylvania Gov. Josh Shapiro and backed by Maryland Gov. Wes Moore.
The Chinese-linked accounts didn't invent that grievance. They found it and poured gasoline on it. That is how influence campaigns work when they are built well enough to be dangerous: they don't need to create a complaint from nothing, they only need to make an existing complaint feel bigger, angrier and more politically useful.
Article 50 of the EU AI Act takes effect August 2, 2026, requiring AI labels on deepfakes, chatbots, and public-interest text across the EU. The European Commission finalized its guidance on July 20 and has expanded its AI Office by 38 staff to monitor firms including OpenAI and DeepSeek, with fines reaching €15 million or 3% of global turnover... - EU AI Act deepfake labeling requirements - AI disclosure law enforcement Europe 2026
The bigger target is permits #
The AI race is now an energy race. Compute is the bottleneck, and compute needs land, transformers, transmission lines, substations and local approvals. If public backlash over electricity prices delays permits, spooks zoning boards or turns state politicians against new builds, that can slow America's AI buildout without touching a server. That is a cheap lever.
This wasn't the first warning. OpenAI said in a June 10 report that it banned a likely PRC-origin cluster of ChatGPT accounts that generated English- and Chinese-language material criticizing U.S. data centers and AI infrastructure. OpenAI said the operators prompted ChatGPT in Simplified Chinese, used VPNs to reach the service, and were likely part of a private Chinese technology company's social media operations team doing work for provincial-level government clients. The campaign had a name inside OpenAI's report: Data Center Bandwagon.
OpenAI's own finding is useful because it cuts against panic. The company said the operation tried to exploit real concerns over energy prices and local data center effects, but found no evidence of meaningful breakout beyond its own activity. Scale is not proof of impact. A giant account farm can still be mostly noise if real people don't pick up the message and carry it for free.
Look at the core. X described 200 active AI and energy-policy accounts inside a 200,000-account farm. That is 0.1% of the network. The company hasn't said how much real engagement those posts received, how the accounts were linked to China, or whether it has evidence tying the operation to a Chinese state agency. No U.S. cybersecurity agency had publicly weighed in on this specific network as of Friday.
The evidence is narrower. X says it found a large coordinated network of suspected Chinese inauthentic accounts and suspended accounts that violated its authenticity policy. OpenAI separately documented a likely PRC-origin operation that used ChatGPT to make anti-data-center posts, comments and cartoons. Together, they show that foreign operators are testing America's AI infrastructure fight as a pressure point.
Frankly, the mistake would be pretending the bots are the whole problem. Data centers really do use large amounts of electricity, and ratepayers really are asking who pays for new grid capacity. Local officials are the ones facing residents who don't want a massive computing site dropped into their community. A foreign influence campaign can exploit that argument, but it can't explain it away.
X's disclosure still matters because it shows where the next pressure campaign may aim: not just elections or tariffs, but the physical infrastructure behind AI itself, where every new facility needs a place on the map and a place on the grid. The fight is still running, one permit hearing and one AI-generated cartoon at a time.
Larry Ellison staked Oracle on OpenAI and now Wall Street is asking what happens if he's wrong
Oracle's S&P downgrade to one notch above junk, a 64% stock collapse from its September 2025 peak, and $117 billion in outstanding bonds have made Larry Ellison the AI boom's most exposed billionaire. Half of Oracle's $638 billion contract backlog runs through a single unprofitable counterparty: OpenAI. - what if oracle's openai bet fails - oracle credit downgrade openai exposure
Also read: Sanctuary AI Will Sell Its Robot Brain, Not Just Its Phoenix Humanoid • Hacker Shows Anyone Near a Unitree G1 Robot Can Seize Root Control • Sony Music and Warner Chappell Sue Anthropic Over Stolen Song Lyrics