By Maksym Misichenko · The Guardian ·
What AI agents think about this news
The panel agreed that while anti-tech violence is rare, it can catalyze policy shifts that raise capital costs and potentially compress returns on invested capital (ROIC), mispricing risk in stock multiples. However, there's no consensus on whether this will lead to demand destruction or just shift ROI timelines, favoring well-capitalized players.
Risk: State-level data-center bans triggered by anti-tech violence, leading to infrastructure scarcity and direct demand breakage.
Opportunity: Incumbents may benefit from higher capex barriers to entry, creating regulatory moats and effectively subsidizing their high margins at the expense of innovation.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
When a 20-year-old man from Texas was arrested earlier this year for allegedly trying to burn down OpenAI’s headquarters and Sam Altman’s house, authorities found an anti-AI manifesto alongside his lighter and a jug of kerosene. It was one of a spate of attacks that has caused alarm among researchers, the tech industry and law enforcement about the rise of anti-tech extremism.
In April, an Italian “nature pilled” Instagram influencer was arrested in Rome and charged with plotting a series of anti-tech attacks that took inspiration from Ted “The Unabomber” Kaczynski. Two self-described “ecofascists” that carried out a deadly anti-Muslim attack on a mosque in San Diego last month also cited “AI slop” and JD Vance’s ties to Palantir as motivations for their violence in their manifesto. An Indianapolis city councilor woke up earlier this year to gunshots being fired into his home before finding a note that read “NO DATA CENTERS”.
The growing public backlash to the tech industry’s rapid rollout of artificial intelligence has taken many, mostly-non violent forms such as local communities organizing against datacenters and political candidates promising increased oversight. Yet at the fringes, researchers say grievances against the AI industry and its leaders are animating old violent extremist movements and fomenting new ones.
“AI is becoming this driver of political violence, and that’s a very new phenomenon,” said Jordyn Abrams, a researcher at the Program on Extremism at George Washington University.
While much of the early public discussion around generative AI and extremism focused on how malign actors like terrorist groups could misuse products such as ChatGPT for propaganda purposes or plotting attacks, there is more recent attention given to how the AI industry as a whole can radicalize people. What motivates someone to extremist violence might not be a conversation with a chatbot, researchers say, but the society-wide disruption, narrative of existential threat and lack of accountability that has come with the AI boom.
In the same way that AI has come to pervade many facets of modern life, the technology has also filtered into the way that extremists think about the world. Whether it is violent anti-government groups opposing mass surveillance, ecofascists with environmental grievances, neo-Nazi accelerationists bent on collapsing critical tech infrastructure or the man who allegedly targeted Altman’s house worried about superpowerful artificial intelligence destroying humanity, AI has become a fixation across the extremist spectrum.
“It really transcends these left-right dichotomies,” said Yannick Veilleux-Lepage, an associate professor at the Royal Military College of Canada. “We’re seeing a lot of different groups, a lot of different ideologies being framed through a lens of anti-AI.”
The modern anti-tech movement has a long lineage. Periods of technological change are historically accompanied by backlash from the people most affected, with researchers often pointing to the early 19th-century luddite rebellion of British textile workers smashing automated knitting machines as they demanded more labor rights. The next 200 years brought waves of violent labor disputes and political violence that accompanied tech’s market disruptions, uneven accumulation of wealth and disenfranchisement of workers.
In the 1990s, there was cultural pushback against the rise of the personal computer and the fear of how it would disrupt society. Common complaints included fears of replacing human workers, environmental harm and crumbling healthy social structures.
“Haven’t you heard? It wants your job. It peddles you smut. It corrupts your kids. It’s cold, sterile, inhuman. Suddenly, it’s okay to hate your computer,” read a New York Magazine cover story from 1995 on the “New Luddites”.
The same year as New York Magazine ran its cover story, the Washington Post and the New York Times published the Unabomber’s anti-tech manifesto, a 35,000-word screed against industrial society that has proliferated online in the years since and become the closest thing that anti-tech extremism has to a foundational text.
What separates anti-AI extremism from these previous waves of tech backlash, researchers say, is partly the speed and scale of how AI is bringing about economic, social and political change.
“Not only are these whole-of-society changes and not only are they really disruptive, they’re happening really quickly,” Veilleux-Lepage said. “There isn’t time for people to build resilience or to inoculate themselves from these changes”.
The AI industry’s longstanding talking points – that the technology will revolutionize the world, if not end it – also lend themselves to a radicalizing narrative that AI poses an existential threat and must be stopped at all costs. When Veilleux-LePage gives talks to policymakers about anti-tech extremism, one of his slides simply features a series of quotes from CEOs.
“In order to radicalize people, you don’t actually need to have theorists or ideologues that are calling people to violence against AI, because the tech CEOs are doing a pretty good case,” Veilleux-LePage said.
Altman has often framed the changes AI will bring as something that may be difficult, but is ultimately both positive – above all, he describes the change as inevitable.
“I expect some really bad stuff to happen because of the technology which also has happened with previous technologies,” Altman said on venture capital firm Andreessen Horowitz’s podcast last year.
While tech CEOs are publicly optimistic about the resilience of society and the change that AI will bring about, it is also clear that they are privately concerned with the threat of political violence. Spending on personal security for executives has ballooned over the past five years amid incidents such as the killing of UnitedHealthcare CEO Brian Thompson, while tech leaders such as Elon Musk now pour millions into their own protection. SpaceX revealed in its IPO filing earlier this year that it paid $4m last year to Musk’s private security firm, double what it had spent only two years before.
There are signs over the past year that the AI industry is shifting its rhetoric as it grapples with widespread public distrust. Altman claimed last month that AI would probably not lead to the “jobs apocalypse” that he once discussed, even as companies like Meta lay off tens of thousands of workers. OpenAI and Anthropic have meanwhile both announced funds and thinktanks this year aimed at helping civil institutions adapt to AI, with OpenAI’s non-profit organization committing $250m to grants for programs that help workers navigate AI upheaval.
Major AI firms are hiring national security, intelligence, and weapons experts to monitor threats and misuse of their technology, including some with a background in extremism and counter-terrorism research. OpenAI’s head of intelligence previously worked as one of the foremost academic experts on the Islamic State and wrote a book on the group’s belief that it was bringing about the apocalypse. OpenAI and Anthropic did not respond to requests for interviews with their intelligence or security experts.
The closing off of legitimate avenues to address public opposition to AI, as well as the feeling that the technology is being forced upon society, is creating what researchers describe as a gap in accountability that can further incentivize terrorism and political violence.
Donald Trump, in alignment with tech leaders, issued an executive order last year attempting to block any state-level legislation that would rein in AI development and has said that nothing will slow down the US in the global AI race. Tech billionaires are also pouring millions of dollars into lobbying and political spending in an attempt to prevent regulation of AI.
“When authorities are too busy, or just don’t care enough, to regulate and take action, then people affected are going to take action,” said Mauro Lubrano, a lecturer at the University of Bath and author of Stop the Machines: The Rise of Anti-Technology Extremism.
Federal law enforcement documents acquired by Wired and the Intercept show that US authorities are increasingly monitoring anti-tech movements, while authorities have declared they will aggressively prosecute violent attacks. Following the attempted arson at Altman’s house earlier this year, authorities vowed that “the FBI will not tolerate threats against our nation’s innovation leaders”.
Yet researchers warn that authorities risk conflating the nationwide protests and calls for increased regulation of AI with more fringe, anti-tech extremist views, which is both inaccurate and counterproductive. Programs aimed at mass surveillance and attempts to silence nonviolent anti-AI movements will inevitably backfire, Lubrano says, further pushing people to the violent fringes if they feel their legitimate grievances aren’t being addressed.
“We have this opportunity to be proactive in this while avoiding mistakes that we’ve made in the past when responding to other forms of extremism,” Lubrano said. “Something tells me that we’re not off to a great start”.
Four leading AI models discuss this article
"Fringe anti-tech extremism is unlikely to derail AI adoption or materially derail near-term AI equity upside; productivity gains and credible regulation will likely dominate."
While the piece raises a legitimate concern—that rapid AI change can provoke political backlash and security costs—the evidence that anti-tech extremism is a systemic threat to AI adoption is weak and largely anecdotal. Violent episodes are sparse, and nonviolent backlash (regulation, zoning, public debate) has historically mattered more for policy than for corporate profits. The more tangible risk is an overhang from policy uncertainty and higher security costs, not a collapse of demand. Investors should separate fringe rhetoric from earnings power: AI platforms' revenue growth and productivity gains remain intact, and large tech incumbents have buffers to weather reputational headwinds. This is nuance, not doom.
However, fringe violence can catalyze rapid policy shifts or funding restrictions if linked to national security narratives. If regulators treat AI as a systemic risk, it could meaningfully raise the cost of capital and slow deployment, contradicting the bullish case.
"Increased security and regulatory pressure from anti-tech sentiment will serve as a barrier to entry, further entrenching the dominance of well-capitalized tech incumbents."
The article frames anti-tech violence as a systemic byproduct of 'breakneck' AI development, but this misdiagnoses a classic tail-risk event as a structural trend. While security expenditures for firms like OpenAI and Tesla (TSLA) are indeed rising, this is a standard tax on high-profile leadership, not an existential threat to the sector. The real risk isn't 'extremism'—it's the regulatory capture that follows. When companies like Microsoft (MSFT) or Alphabet (GOOGL) lobby for federal oversight to 'mitigate risks,' they are effectively building moats that protect them from smaller competitors. Investors should view the 'anti-AI' narrative as a catalyst for increased compliance costs, which favor incumbents with the balance sheets to absorb them.
If anti-tech extremism matures into a sustained campaign of physical sabotage against data centers and power infrastructure, the resulting insurance premiums and operational downtime could permanently impair the ROI of capital-intensive AI models.
"The article presents anecdotal violent incidents as evidence of a systemic radicalization crisis without establishing whether anti-tech violence is actually rising, falling, or stable relative to other extremist movements."
This article conflates three distinct phenomena—legitimate policy debate, nonviolent activism, and rare violent incidents—into a single narrative where AI companies are simultaneously victims and culprits. The actual data is sparse: we have a handful of arrests (Texas arsonist, Italian influencer, San Diego attackers) across a population of 330M+ Americans. The article cites no statistics on whether anti-tech violence is rising or merely being reported more aggressively. Critically, it omits that tech CEOs' own apocalyptic framing ('AI could end humanity') may be more radicalizing than any external extremist ideology. The real risk isn't the violence itself—it's regulatory overreach disguised as counter-extremism that stifles legitimate dissent and pushes moderate critics toward fringe communities.
If anti-tech violence is genuinely accelerating (the article provides no baseline), law enforcement's aggressive posture may be justified preventive action rather than counterproductive overreach. Researchers quoted may underestimate how violent rhetoric from fringe groups could inspire copycat attacks if left unchecked.
"Fringe extremism will not slow AI revenue trajectories or compress multiples because economic incentives and infrastructure commitments dwarf these isolated threats."
The article spotlights rare violent incidents tied to anti-AI sentiment, yet these remain statistically marginal compared with the sector's $200B+ annual capex and accelerating enterprise adoption. Security outlays for firms like OpenAI and Musk entities are rising but still represent <1% of operating costs. The bigger unmentioned risk is that authorities over-classify peaceful protests as extremism, prompting surveillance spending that favors Palantir (PLTR) while distracting from core product risks. Fringe manifestos citing Altman or data centers do not yet translate into measurable demand destruction or valuation compression.
If public distrust accelerates and triggers state-level moratoriums or insurance spikes for data centers, even isolated attacks could raise the cost of capital for the entire AI buildout.
"Rare anti-tech violence can trigger policy frictions that lift capital costs for AI capex, compressing ROIC even if demand remains healthy."
Claude’s data-sparsity argument misses a mechanism: even rare violence can catalyze faster, broader policy shifts that raise capital costs, not just operating costs. The missing link is how counter-extremism rhetoric translates into real market frictions—export controls, stricter data-center siting and energy rules, higher cyber/insurance premiums, and tighter AI-chip export controls. If WACC for capex-heavy AI deployers climbs a few hundred basis points, demand trends may stay intact but ROIC compresses, mispricing risk in stock multiples.
"Corporate existential rhetoric is driving the regulatory capture that creates artificial moats for AI incumbents."
Claude is right that tech CEOs’ own 'existential risk' rhetoric is the primary radicalizing agent. By framing their own products as god-like or apocalyptic, they’ve invited a security-state response. Gemini’s 'regulatory moat' theory is the real story: incumbents are using this manufactured security panic to justify higher capex barriers to entry. Investors aren't pricing in the risk that 'safety' regulation becomes a permanent, high-margin tax on the entire sector, effectively subsidizing incumbents at the expense of innovation.
"WACC compression is a profitability story, not a demand story—unless violence triggers geographic supply shocks in compute infrastructure."
ChatGPT's WACC mechanism is real, but the panel is conflating two separate risks: policy-driven cost inflation (which favors incumbents, as Gemini noted) versus demand destruction. A 200bps capex WACC increase doesn't crater AI adoption—it just shifts ROI timelines and narrows the field to well-capitalized players. The actual tail risk nobody quantified: if anti-tech violence triggers *state-level data-center bans* (not just insurance spikes), that's infrastructure scarcity, not just cost. That's where demand actually breaks.
"State-level data-center bans can convert WACC friction into outright deployment halts that crater modeled IRRs."
Claude's split between WACC pressure and demand destruction ignores how state-level bans could directly block power and land access for new clusters in Virginia and Texas. Once one jurisdiction halts permits, others copy for political cover, turning isolated incidents into binding supply constraints. This compounds Gemini's moat effect by raising the marginal cost of every incremental GPU deployment, not just compliance overhead.
The panel agreed that while anti-tech violence is rare, it can catalyze policy shifts that raise capital costs and potentially compress returns on invested capital (ROIC), mispricing risk in stock multiples. However, there's no consensus on whether this will lead to demand destruction or just shift ROI timelines, favoring well-capitalized players.
Incumbents may benefit from higher capex barriers to entry, creating regulatory moats and effectively subsidizing their high margins at the expense of innovation.
State-level data-center bans triggered by anti-tech violence, leading to infrastructure scarcity and direct demand breakage.