AI super PAC sockpuppets: the fake doomer scandal, explained
The AI super PAC sockpuppet story reveals how pro-AI political money, fake personas, data centers and local AI reshape data center politics.
The viral claim sounds like something designed to melt your brain: a pro-AI super PAC network supposedly paid for fake anti-AI “doomer” accounts, including a staged persona called Jonathan Doomer, to make opposition to Big Tech AI and data centers look deranged, violent and pro-China.
The evidence is serious enough to investigate. It is not strong enough to compress into a viral line like “OpenAI ran fake death-threat accounts.” The real story is more useful, and more disturbing, because it shows how AI politics is becoming a money war over public legitimacy, local infrastructure and who gets to define dissent.
At the center of the story is a simple question: when AI companies need data centers, favorable regulation and public permission, how far will their political ecosystem go to make critics look unreasonable?
Key takeaways
The real political machine exists. Leading the Future is an active federal super PAC registered with the FEC, and Build American AI describes itself as associated with Leading the Future and as a 501(c)(4) policy force for a pro-AI agenda.
The money is real. The FEC lists Leading the Future as an independent expenditure-only super PAC with $75.5 million in receipts through March 31, 2026.
The influence campaign is real. WIRED reported that Build American AI funded a paid influencer campaign to promote pro-AI messaging and frame Chinese AI as a threat.
The sockpuppet claim is reported, but the public evidence is mixed by layer. User Mag and The Midas Project reported that pro-AI meme and fake anti-AI accounts were linked to the Leading the Future ecosystem. Model Republic separately reported circumstantial evidence around another anonymous account, Doomer Daylight, and explicitly said it had not seen a contract or email proving who operated it.
OpenAI says it has not donated to super PACs, does not direct Leading the Future, and no outside political group speaks for the company.
The bigger issue is the control lever. Big AI needs data centers, federal preemption, political protection, and social permission. Local, decentralized, open-source AI weakens that chokepoint by giving users more capability outside cloud accounts and hyperscale infrastructure.
What people are claiming online
The most viral version of the claim goes roughly like this: a pro-AI super PAC tied to OpenAI and Andreessen Horowitz secretly paid for fake social media accounts on X. One side of the operation mocked AI critics. The other side pretended to be an unhinged anti-AI doomer who praised China, exaggerated AI apocalypse fears and posted violent or threatening rhetoric about AI executives and data centers.
According to that version, the goal was to make real opposition to Big Tech AI look crazy, foreign-aligned and dangerous.
That is the strongest version of the allegation. It travels well online because it is simple, explosive and emotionally satisfying. It also needs to be handled carefully.
The evidence-backed version is narrower.
A pro-AI political network called Leading the Future exists. It has serious money behind it. It has an affiliated 501(c)(4), Build American AI. WIRED documented a paid influencer campaign funded by Build American AI. User Mag and The Midas Project reported that multiple meme and persona accounts, including “Doomers Are Dumb” and “Jonathan Doomer,” appeared linked to that political network or its vendors. Model Republic separately reported circumstantial evidence connecting another anonymous account, “Doomer Daylight,” to the same broader ecosystem.
That evidence does not prove every viral inference. It does show that AI politics has already moved into paid narrative warfare.
The real pro-AI political machine
The Federal Election Commission lists Leading the Future as an active, unauthorized, independent expenditure-only super PAC, registered on August 15, 2025. The FEC’s committee profile reports $75,535,801.59 in total receipts and $24,390,931.31 in total disbursements for the period from August 15, 2025, through March 31, 2026.
That is not a loose activist group, a fan account or a rumor. It is a registered federal political committee with tens of millions of dollars moving through it.
Leading the Future’s own launch release said the operation was backed by more than $100 million, with named supporters including Andreessen Horowitz, Greg and Anna Brockman, Ron Conway, Joe Lonsdale and Perplexity. The same release said the network would include federal and state super PACs plus 501(c)(4) issue advocacy groups, giving it flexibility across the 2026 election cycle.
That structure matters because different entities leave different paper trails.
A super PAC must disclose more than a 501(c)(4). A 501(c)(4) can do issue advocacy with less public visibility into donors and spending. If political work happens through the c4 layer, or through a vendor hired by that layer, the public record becomes much harder to follow.
Build American AI makes the connection plain on its own site. In its candidate questionnaire, Build American AI says it is associated with Leading the Future and describes itself as a central policy force for a pro-AI agenda. The questionnaire asks candidates whether they support a national AI framework rather than state-by-state rules, whether the federal government should help AI innovators build facilities and manage infrastructure impacts, and whether domestic energy and infrastructure should support the AI buildout.
That is the official policy frame: national AI leadership, China competition, federal consistency, infrastructure buildout and support for AI companies.
The question is what happens when that frame runs into public backlash.
Why data centers are central to the story
The AI super PAC sockpuppet story is not only about online drama. It is about physical infrastructure.
Cloud AI needs enormous data centers. Those data centers need land, water, power, transmission, local permits, tax incentives, political allies and a public story that makes the buildout feel necessary.
The public is not automatically buying that story. Gallup reported in May 2026 that seven in 10 Americans oppose constructing AI data centers in their local area. Gallup found that opposition was tied to concerns about water use, energy use, pollution, local quality of life, utility bills and cost-of-living pressure.
That makes data centers a political problem for Big AI. It also explains why the “crazy anti-AI doomer” frame is so useful. If opposition can be made to look irrational, violent, foreign-aligned or anti-progress, local resistance to data centers becomes easier to dismiss.
This is where the online sockpuppet allegation connects to zoning meetings, grid capacity, utility bills and local organizing. A fake persona on X may look small. A narrative that turns data center critics into extremists can be much larger than the account that launches it.
Build American AI’s questionnaire points to the same pressure point. It asks candidates to agree that the federal government should help AI innovators build new facilities, navigate red tape and mitigate local infrastructure impacts. In plain English, that means data center politics.
The paid influencer campaign is documented
The strongest evidence of organized narrative shaping comes from WIRED, not from screenshots of anonymous accounts.
In May 2026, WIRED reported that Build American AI funded a paid influencer campaign to spread pro-AI messaging and frame Chinese AI as a threat. WIRED said marketing agencies pitched influencers deals such as $5,000 per TikTok video, and that the campaign’s first phase used lifestyle influencers to promote American AI innovation while the second phase emphasized China.
That matters because it proves the ecosystem was willing to pay for social media narrative work. It does not prove the sockpuppet allegation on its own. It does make the allegation more plausible than it would be in isolation.
There is a difference between paid influencer messaging and sockpuppetry.
An influencer post can be disclosed as an advertisement, even when the disclosure is imperfect. WIRED reported that one influencer labeled a post as an ad without naming the paying organization. A sockpuppet account is more deceptive because it pretends to be an independent person, faction or movement.
The viral allegation is about that second category. It is about fake identity, not ordinary messaging.
The sockpuppet reporting
The central sockpuppet article is Taylor Lorenz and Tyler Johnston’s User Mag report, “A Pro-AI Super PAC’s Secret Meme Sockpuppets”. The report says an OpenAI and a16z-affiliated pro-AI super PAC appeared linked to multiple sockpuppet accounts that attacked or parodied critics with offensive memes.
The two accounts that drove the viral discussion were “Doomers Are Dumb” and “Jonathan Doomer.”
According to the reporting summarized publicly by The Midas Project and related social posts, the claimed pattern was that “Doomers Are Dumb” functioned as a pro-AI meme account mocking AI critics, while “Jonathan Doomer” presented as an AI skeptic or anti-AI persona. The fake critic allegedly made AI opposition look absurd, extreme, pro-China or threatening. The accounts had small organic followings, which makes their importance less about reach and more about the tactic.
That distinction matters. The story is not that these accounts were major influencers in their own right. The story is that they may have supplied a caricature. A fake extremist does not need millions of followers to be useful. It only needs to provide screenshots, vibes and proof-shaped material that can be pointed at later.
The reporting connected the alleged operation to Memelord Technologies and to the Leading the Future or Build American AI ecosystem.
The most important caution is that much of this evidence is still available publicly as reporting, snippets, screenshots, posts and commentary rather than as a clean official admission, authenticated contract or court-tested record.
The earlier Doomer Daylight case
Before the Jonathan Doomer story, Tyler Johnston at Model Republic published a detailed investigation into another anonymous account, “Doomer Daylight.”
That piece alleged the account was likely connected to the Leading the Future ecosystem. The evidence included shared web infrastructure, overlapping privacy policy language, similar posting patterns, shared messaging, paid promotion and personnel links to Targeted Victory, the political consulting firm associated with Leading the Future’s leadership.
The most credible part of that investigation is also its most important limitation. Johnston wrote that he had not seen a contract or email proving Build American AI or Targeted Victory operated the account. He described the evidence as a convergence, not a smoking gun.
That is the correct evidence tier for much of this story: suspicious, relevant and worth reporting, but not the same as a legal finding.
The Doomer Daylight case also shows how these influence stories often work. They rarely arrive as one perfect document that proves everything. They arrive as domain registrations, vendor overlap, web templates, ad behavior, account timing, reused language and posting patterns.
What is proven by public records
Leading the Future exists as a federal super PAC. The FEC lists it as active, quarterly filing, unauthorized and independent expenditure-only.
Leading the Future has raised major money. FEC data shows more than $75 million in receipts through March 31, 2026.
Leading the Future publicly announced more than $100 million in initial funding and named supporters including Andreessen Horowitz and Greg and Anna Brockman.
Build American AI describes itself as associated with Leading the Future and as a 501(c)(4) supporting candidates who champion an AI-forward future.
Build American AI’s candidate questionnaire pushes a national AI framework, federal partnership with AI innovators, and energy and infrastructure support for AI.
WIRED reported that Build American AI funded a paid influencer campaign using pro-AI and anti-China messaging.
OpenAI says it has not made super PAC donations, does not have an employee-funded PAC, has not donated to candidates or campaigns, and does not direct Leading the Future.
What is reported but still incomplete
User Mag and The Midas Project reported that pro-AI meme and fake anti-AI accounts were linked to the Leading the Future ecosystem.
Model Republic reported circumstantial evidence linking the Doomer Daylight account to the same broader ecosystem.
Public summaries of the User Mag report describe Jonathan Doomer as a fake AI-critical persona whose posts made AI skeptics look extreme, including through pro-China rhetoric and violent or threatening language.
Guardian reporting shows AI super PAC spending is now a major factor in congressional races, including the NY-12 race involving Alex Bores, sponsor of New York’s RAISE Act.
WIRED reported that law enforcement agencies and fusion centers have started circulating warnings around a broad “anti-technology extremism” frame, including data center opposition.
These claims are serious. They should be described as reported claims unless stronger public documentation appears.
More on corporate AI:
What is still unproven
There is no public record reviewed here proving that OpenAI corporate management ordered, paid for or directly operated Jonathan Doomer.
There is no public court filing, regulator finding or authenticated contract reviewed here proving that Build American AI hired a vendor to operate Jonathan Doomer.
There is no public evidence reviewed here proving that fake doomer posts caused law enforcement surveillance of data center critics.
There is no basis to treat all AI critics, AI safety advocates, data center opponents, local residents or local AI users as violent extremists.
That last point matters. Fringe threats do not erase legitimate opposition to data centers, job displacement, surveillance, platform control or AI regulation capture.
OpenAI’s denial and why it matters
OpenAI published a June 1, 2026 statement on AI policy and political advocacy. The company said it has not donated to any super PACs, does not have an employee-funded PAC, and has not donated to political candidates or campaigns.
OpenAI acknowledged that Leading the Future received support from OpenAI president and co-founder Greg Brockman and his wife Anna, but said that support was in a personal capacity. OpenAI also said it does not direct Leading the Future or have visibility into its operations, and that no outside political group speaks for OpenAI.
The same statement says groups advocating on AI should be honest about whom they represent and should avoid astroturfing tactics that obscure real policy choices.
That statement is important for two reasons.
First, it is a denial of corporate direction.
Second, the astroturfing line creates a standard. If a pro-AI political network uses fake personas, anonymous paid accounts or undeclared influence campaigns, the problem goes beyond bad optics. It becomes an attack on public accountability in AI policy.
The NY-12 test case
The Guardian reported on June 22, 2026, that AI-focused super PACs have raised roughly $100 million in the 2026 cycle, with about $44 million spent so far, and nearly half of that spending going to New York’s 12th congressional district.
That race matters because Alex Bores sponsored New York’s RAISE Act, a state-level AI safety law requiring major AI developers to publish public safety plans. The Guardian reported that Think Big, an affiliate of Leading the Future, poured $8.2 million into the primary.
This is the political version of the same control fight.
If state-level AI rules are framed as a dangerous patchwork, Big AI can push for a federal framework. If local data center resistance is framed as irrational, federal and state politicians can justify faster infrastructure approvals. If AI critics are framed as doomer extremists, elected officials have more incentive to keep them at a distance.
The control lever is legitimacy.
Who gets to be seen as reasonable?
Who gets labeled a threat?
Who gets invited into policy rooms?
Who gets turned into a meme?
Those questions are not abstract. They decide who has power when AI policy moves from Twitter arguments into elections, utility boards, local permitting and federal law.
The law enforcement angle
The sockpuppet allegation becomes more serious when placed next to the emerging “anti-tech extremism” frame.
WIRED reported in May 2026 that federal intelligence agencies, the FBI, DHS and fusion centers circulated reports around a broad new category of anti-technology extremism. WIRED’s reporting was based on more than 1,000 pages of documents obtained through public records requests.
Some reports focused on real risks of violence or sabotage. That matters. Threats against executives, workers, facilities and infrastructure should be treated seriously.
The danger is category creep. WIRED reported that some suspicious activity indicators could be carried out by peaceful protesters, and that fusion centers had circulated intelligence about civic events and demonstrations related to critical views of technology. WIRED also reported that a More Perfect Union video about data center harms was flagged through this kind of monitoring, even though the video did not advocate violence.
This is where fake extremist personas, if proven, would be especially corrosive.
A fake violent “doomer” account does not need to convince many people. It only needs to supply screenshots, talking points or emotional confirmation for people already eager to merge legitimate AI opposition with extremism.
The control lever: narrative laundering
The central control lever in this story is narrative legitimacy.
Big AI already has several obvious control points: cloud accounts, API access, model weights, app store placement, content policy, data center buildout, compute supply, political spending, federal preemption, paid media and influencer campaigns.
The sockpuppet allegation points to a softer control point: the power to define the opposition.
If a political operation can create or amplify the ugliest possible caricature of AI criticism, it can shift the debate away from real questions.
Should local communities bear the power and water costs of data centers?
Should state governments be blocked from regulating frontier model developers?
Should AI policy be written around the needs of hyperscale labs?
Should open models and local inference be treated as public goods or risks to be contained?
Should ordinary users depend on a few cloud platforms for work, search, speech, coding and media creation?
Those are serious policy questions. They are also inconvenient questions for companies and political groups that want the AI buildout to move quickly.
A caricature is easier to defeat. The anti-AI doomer can be painted as unstable, anti-progress, violent, pro-China and unserious. That version of the critic gives politicians permission to ignore the local resident worried about water, the worker worried about automation, the open-source developer worried about licensing capture and the privacy advocate worried about cloud dependency.
That is useful if the political goal is to make Big AI’s critics radioactive.
Why this is bigger than doomers versus accelerationists
This story is often framed as “doomers versus accelerationists.” That frame is too small.
Some AI safety advocates want central regulation that could strengthen incumbents. Some accelerationists want open models and user-owned compute. Some Big Tech companies use safety language when it protects them. Some anti-regulation groups use open innovation language while pushing cloud infrastructure and political centralization.
The useful dividing line is different: does the policy increase user capability and competition, or does it consolidate control?
A national AI framework can help or hurt depending on what it does. If it prevents 50 contradictory rules and protects open-source developers, it may help. If it preempts states while giving hyperscale companies the easiest compliance path, it becomes a moat.
AI safety advocacy can help or hurt depending on its mechanism. If it creates transparency, liability for real harm and public accountability, it may help. If it creates licensing systems that only the richest labs can survive, it becomes capture.
Pro-AI advocacy can help or hurt depending on its honesty. If it argues openly for faster innovation, that is persuasion. If it hides behind fake people, undeclared influencers or caricatured opposition, that is manipulation.
The danger is that both sides can end up strengthening central control. One side can argue that AI is too dangerous for ordinary builders. The other can argue that only hyperscale companies can move fast enough to beat China. Both paths can leave users more dependent on a few companies, a few cloud platforms and a few political narratives.
Why local AI changes the political equation
Cloud AI and local AI are not merely technical choices. They create different political incentives.
Cloud AI centralizes capability in a few companies with massive infrastructure needs. Those companies need data centers, energy deals, water access, transmission lines, chip supply, favorable regulation, friendly procurement and public permission. They also control accounts, logs, model access, safety filters, pricing and API terms.
Local AI distributes capability. A local LLM running through Ollama, LM Studio, llama.cpp, GGUF Loader, Open WebUI or a similar stack is weaker than the best frontier model for many tasks. It can still be enough for private writing, coding help, summarization, search, notes, transcription, retrieval and routine automation.
That matters because “AI progress” does not have to mean more dependence on a few cloud companies.
Popular AI has already covered why users should keep a local AI fallback even when frontier models improve, how to evaluate whether local AI hardware is worth buying in 2026, and how to compare llama.cpp, Ollama and LM Studio. For private workflows, a local path is not always faster. It can still reduce account risk and keep more data on hardware the user controls.
This is the practical answer to both sides of the AI politics trap.
Do not let Big AI turn every criticism into anti-technology panic.
Do not let AI doom politics turn every powerful model into an excuse for centralized licensing.
Build capability that does not require permission.
What local AI cannot solve
Local AI is not a magic shield against politics.
It still depends on GPUs, drivers, open-source maintainers, model licenses, chip supply, electricity and the broader software ecosystem. Local models can hallucinate. Open-weight licenses can include restrictions. Consumer hardware cannot match every cloud frontier model. Training large models is still out of reach for ordinary users.
Local AI also does not eliminate data centers. Frontier research, enterprise inference, video generation, scientific workloads, robotics and large-scale services still require serious compute.
But local AI changes the default bargain.
Instead of accepting “AI equals hyperscale cloud,” users can split workloads. Cloud AI can be the right choice when frontier reasoning, speed or tool integration matters. Local AI can be the right choice for private files, offline work, routine automation, local search, transcription, draft generation, coding helpers and experiments.
The point is not ideological purity. It is practical resilience.
Keep exportable data. Avoid building an entire business process around one hosted account. Learn enough about local inference to keep leverage. The more users can run, inspect, move and own, the less the AI debate has to revolve around the needs of a few cloud companies.
More on the pros and cons of local AI:
How readers should evaluate the next viral claim
When the next viral AI astroturfing claim appears, use a simple evidence ladder.
Public records are the strongest starting point. Ask what is documented by FEC filings, corporate statements, court records, official campaign pages, ad libraries, tax filings or contracts. In this case, the existence of Leading the Future, its FEC status and its disclosed receipts are public record.
On-record reporting comes next. Ask whether named reporters or outlets reviewed documents, obtained direct confirmations, contacted targets for comment and linked sources. WIRED’s influencer story is strong because it describes direct outreach, agency pitches, confirmation from multiple creators and a comment from a spokesperson representing Leading the Future.
Technical and behavioral evidence can also matter. Infrastructure overlap, posting patterns, shared language, account timing, mutual engagement and ad behavior can be persuasive. That kind of evidence is still different from a contract.
Screenshots and social summaries should be treated as leads. They can be real, fake, cropped, deleted or missing context. They can help identify what to investigate. They should not become the final proof.
Viral certainty is the lowest tier. When a thread turns “reported links” into “Company X did Y,” slow down. The viral version is often emotionally satisfying and legally sloppy.
The best analysis of this subject should be harder to dismiss than the propaganda it criticizes.
FAQ
Did OpenAI run the Jonathan Doomer account?
The public evidence reviewed here does not prove that OpenAI the company ran Jonathan Doomer. OpenAI says it has not donated to super PACs, does not direct Leading the Future, and no outside political group speaks for the company. The reported connection is to a pro-AI political ecosystem backed by tech figures, including OpenAI president Greg Brockman and his wife in a personal capacity, plus a16z-linked funding.
Was there really an AI super PAC?
Yes. Leading the Future is listed by the FEC as an active independent expenditure-only super PAC. Its launch release described a broader network of super PACs and 501(c)(4) advocacy organizations.
What is Build American AI?
Build American AI is a 501(c)(4) advocacy group associated with Leading the Future. Its candidate questionnaire describes it as a central policy force for a pro-AI agenda and asks candidates about national AI rules, China competition, federal support for AI infrastructure, energy, jobs, and AI in public services.
What is the strongest confirmed influence operation?
The strongest confirmed influence operation is the paid influencer campaign reported by WIRED. WIRED said Build American AI funded social media messaging that promoted American AI and framed Chinese AI as a threat.
What is the strongest sockpuppet evidence?
The strongest public sockpuppet evidence is investigative reporting by User Mag, The Midas Project, and Model Republic. Model Republic’s Doomer Daylight article is especially useful because it lays out technical and behavioral evidence while admitting that it has not seen a contract or email proving operation of the account.
Did fake AI doomer accounts threaten AI CEOs?
Viral summaries say the fake doomer persona posted violent or threatening rhetoric about AI executives and data centers.
Why would a pro-AI group create a fake anti-AI account?
If the allegation is true, the likely purpose would be caricature. A fake anti-AI persona can make critics look irrational, violent, anti-American, or pro-China. That helps a pro-AI political operation argue that it represents seriousness, jobs, progress, and national security.
Is opposition to AI data centers fringe?
No. Gallup found in May 2026 that seven in 10 Americans opposed AI data centers in their local area. The reasons included water use, energy use, pollution, quality of life, utility bills, and local costs.
Does this prove AI regulation is good?
No. The story proves that AI politics is becoming an influence war. Some regulation could increase transparency and accountability. Some regulation could also protect incumbents and squeeze open-source or local AI builders. The question is always the mechanism.
Does this prove accelerationists are bad?
No. It proves that political money and anonymous influence campaigns need scrutiny, regardless of whether they are pro-AI, anti-AI, safety-branded, or accelerationist. The useful split is honest capability versus centralized control.
TL;DR
Treat the AI super PAC sockpuppet story as a warning about power, not as an excuse for sloppy claims.
The strongest supported conclusion is this: Big AI’s political infrastructure is now large, well-funded and willing to shape public opinion through paid media, influencer work, candidate pressure and aggressive narrative framing. Reported sockpuppet activity fits that pattern, but the public evidence does not yet prove every viral claim about OpenAI or every specific account.
The practical response is not to ban AI or surrender AI policy to the same companies building the data centers. It is to demand transparency from political advocacy groups, resist fake grassroots theater, separate real threats from legitimate dissent and build AI capability that does not depend entirely on cloud accounts.
Use frontier cloud AI when it earns its place.
Build local AI where privacy, resilience and control matter.
Do not let either doom politics or Big Tech politics turn AI into a permission system.
Explore more from Popular AI:
Start here | Local AI | Fixes & guides | Builds & gear | Popular AI podcast