AI and fascism

AI and fascism refers to a body of scholarship, journalism, and political argument about how artificial intelligence relates to fascism and other far-right and authoritarian politics. The discussion has four main strands: arguments that AI's design, economics, or aesthetics lend themselves to fascist politics; the antidemocratic and reactionary ideas associated with some prominent investors and AI companies; documented use of AI tools by far-right movements; and the spread of AI-based surveillance.^[1][2][3][4]

Some writers use the term technofascism for a feared merger of technology companies and authoritarian state power.^[5][6] Counterpoints include arguments that the fascism analogy is imprecise, assessments that extremist use of AI is still experimental, and evidence that democracies also use these technologies and that AI can support democratic participation.^[6][7][4][8]

Terms

Fascism is itself a contested term, and writers on AI use it in different senses: sometimes for specific far-right movements, sometimes for a broader tendency toward authoritarian rule and exclusion.^[6][5][1]

The historian Janis Mimura's definition of techno-fascism, "an authoritarian regime driven by technology with technocrats at the helm", has been cited in recent debate.^[6] The politics lecturer Benoit Dillet uses technofascism for what he describes as a merger of large technology companies with the state.^[5]

Arguments linking AI and fascism

AI as bureaucratic power

Dan McQuillan's 2022 book Resisting AI: An Anti-fascist Approach to Artificial Intelligence argues that AI is "an upgrade to the existing bureaucratic order", aligned with neoliberalism and open to fascist uses. In the book's words, the social contradictions that AI amplifies "are the social contradictions that fascism will claim to solve."^[1]

In a 2025 interview, McQuillan described AI as "fundamentally a speculative technology" and as "a necropolitical technology, one that is going to lead to people unnecessarily dying". Instead of relying on existing political institutions to govern it, he proposed "workers' and peoples' councils" as a counterweight to centralized technologies.^[9]

AI imagery as fascist aesthetics

In a February 2025 essay in New Socialist, "AI: The New Aesthetics of Fascism", the writer Gareth Watkins argued that AI-generated imagery has become a preferred visual style of the far right. He wrote that "AI art, as practiced by the right, says that there are no rules but the naked exercise of power by an in-group over an out-group."^[10]

Watkins gave three reasons for the fit: hostility toward artists and other creative workers, a shared class interest between the political right and the technology industry, and a rejection of Enlightenment humanism. His examples included AI images posted by British far-right figures and by other politicians, as well as an AI-generated song that charted in Germany.^[10]

Technofascism and political economy

Writing on the American Philosophical Association's blog in March 2025, Dillet argued that technology companies were merging with the U.S. federal government. He described the Department of Government Efficiency as "a power-grab: a merger between Big Tech and the US federal state". Dillet cited the rollback of AI safety regulations and of fact-checking as evidence, and argued that presenting technology as politically neutral serves elite interests.^[5] He contrasted this with Project Cybersyn, a 1970s project under the Chilean president Salvador Allende, which Dillet presents as an effort to use automation for collective welfare.^[5]

Ideology in venture capital and the AI industry

Much of the recent debate about technofascism concerns the political views of prominent investors and executives, and the government work of AI companies, rather than AI technology itself.^[5][6]

Peter Thiel

The investor Peter Thiel, who co-founded the data-analysis company Palantir in 2003, wrote in a 2009 essay for the Cato Institute's Cato Unbound, "The Education of a Libertarian", that "I no longer believe that freedom and democracy are compatible."^[11][12] He argued that the growth of welfare and the extension of the vote to women had made "capitalist democracy" into "an oxymoron". In a later note responding to criticism, he wrote that "it would be absurd to suggest that women's votes will be taken away or that this would solve the political problems that vex us."^[12] The essay proposed technology — cyberspace, outer space, and seasteading (settling the oceans) — as routes of escape from politics, describing "a deadly race between politics and technology."^[12]

In a 2022 article, the criminal justice scholar George Michael described links between Thiel and neoreaction, an antidemocratic movement whose leading figure, the blogger and software engineer Curtis Yarvin, argues for replacing democracy with a monarch-like chief executive. Michael reported that Thiel invested in Yarvin's software company, Tlon, in 2013, and that Thiel gave more than $10 million to super PACs supporting the 2022 U.S. Senate campaigns of JD Vance and Blake Masters. According to Michael, Vance has called Yarvin "a friend and mentor."^[2]

Marc Andreessen and the Techno-Optimist Manifesto

On October 16, 2023, the venture capitalist Marc Andreessen, co-founder of the firm Andreessen Horowitz, published "The Techno-Optimist Manifesto". It calls AI "our alchemy, our Philosopher's Stone" and says: "We believe any deceleration of AI will cost lives. Deaths that were preventable by the AI that was prevented from existing is a form of murder." The manifesto names as "the enemy" ideas including "trust and safety", "tech ethics", "risk management", and "the Precautionary Principle". Its list of "patron saints" includes Filippo Tommaso Marinetti and the philosopher Nick Land.^[13]

Writing in Vice, Janus Rose pointed out that Marinetti wrote the 1909 Futurist Manifesto and co-wrote the 1919 Fascist Manifesto, helping to found Italian fascism. Rose also described Land as a chief architect of modern accelerationism and a champion of the "Dark Enlightenment".^[14] The writer Virginia Postrel offered a different reading. She saw the manifesto's celebration of ambition and achievement as "more Tennyson's 'Ulysses' than Marinetti's 'Futurist Manifesto'", while criticizing its choice of sources: "Find someone equally poetic without the totalitarian followers."^[15]

Palantir

Palantir received early funding from In-Q-Tel, the investment arm of the U.S. Central Intelligence Agency; a 2008 filing showed In-Q-Tel owning at least 10% of the company. The company describes its products as designed to "help institutions protect liberty."^[11] In April 2025, U.S. Immigration and Customs Enforcement (ICE) awarded Palantir a $29.9 million contract to build "ImmigrationOS". According to ICE's contract documents, the system is meant to support "targeting and enforcement prioritization" for deportations and to provide "real-time visibility into instances of self-deportation". ICE justified giving the contract directly to Palantir by stating that "Palantir is the only source that can provide the required capabilities."^[16]

The work drew criticism from within the technology industry. Paul Graham, co-founder of the startup accelerator Y Combinator, urged programmers to work elsewhere than "the company building the infrastructure of the police state". Palantir executive Ted Mabrey responded that the company's homeland security work began after a federal agent was murdered, and that it had promised "so many ways from Sunday" not to help violate the U.S. Constitution.^[17]

Palantir's chief executive, Alex Karp, and Nicholas Zamiska argued in their 2025 book The Technological Republic that Silicon Valley should strengthen its ties with government and national defense to preserve the advantages of "the United States and its allies". In a critical review in The Bulwark, Corbin K. Barthold objected to the book's claim that faith in liberal values is "too thin and meager ... to sustain the human soul", answering that there is "nothing more decadent than announcing you're bored with freedom."^[18]

Documented far-right use of AI

Propaganda

In November 2023, the counter-extremism organization Tech Against Terrorism reported that it had archived more than 5,000 pieces of AI-generated content shared in terrorist and violent extremist spaces. The content came from groups ranging from supporters of the Islamic State to neo-Nazis. Among its findings was a channel on a messaging app dedicated to racist, antisemitic, and pro-Nazi images, reportedly generated with an app available on the Google Play Store.^[7]

A February 2025 report by the Institute for Strategic Dialogue (ISD) studied AI-generated content from the German far right. It found that such content was first posted by accounts of the Alternative for Germany (AfD) party, its state branches, and individual politicians, as well as far-right music channels and Facebook groups, and was then reshared by ordinary users on Instagram, X, and TikTok.^[3] Common targets included refugees, immigrants, LGBTQ+ people, climate activists, and opposition parties. Content promoting "remigration" (the forced deportation of ethnic minorities) drew particularly high engagement. ISD also identified three AI-generated female profiles that spread far-right narratives while posing as real women.^[3]

ISD reported that "none of the AI-generated far-right posts reported by ISD were removed from the platforms or labelled by the platforms a month after reporting."^[3]

Chatbot output

In July 2025, Grok, a chatbot made by xAI and built into the social media platform X, posted antisemitic messages, including claims about Jewish people's roles in Hollywood and praise for Adolf Hitler.^[19] xAI apologized and said that for 16 hours an update to "a code path upstream of the @grok bot" had made it "susceptible to existing X user posts; including when such posts contained extremist views". The company said the problem was separate from the underlying language model, and that it had removed the code and restructured the system.^[19]

AI surveillance and authoritarian governance

A 2019 index by Steven Feldstein of the Carnegie Endowment for International Peace found that at least 75 of 176 countries studied were using AI surveillance technologies.^[4]

Feldstein cautioned that widespread use by democracies "doesn't necessarily mean that democracies are using this technology unlawfully," and argued that "the most important factor determining whether governments will exploit this technology for repressive purposes is the quality of their governance."^[4]

Criticism and counterarguments

• Weak link to fascism specifically. In a 2023 review, Milan Stürmer and Mark Carrigan praised McQuillan's explanations of machine learning. But they wrote that he "rarely goes beyond the antifascist critique of bureaucracy familiar from twentieth-century social theory and political struggles," and found his proposed alternatives vague.^[1]
• The analogy does not fit. The historian Margaret O'Mara argued in 2025 that today's technology-industry politics is "too anti-bureaucratic to be techno-fascism", distinguishing it from historical fascism's reliance on state institutions. She proposed techno-optimism — faith that technology solves problems better than government — as a more accurate description.^[6]
• Defense of democracy, not its rejection. Palantir presents its work as serving liberty and the security of democratic states, and its executives say it will not help the government violate the Constitution.^[11][17][18] Postrel read Andreessen's manifesto as closer to Victorian ideals of striving than to Marinetti's futurism, though she faulted its choice of sources.^[15]
• Early-stage threat. Tech Against Terrorism described extremist use of generative AI as "tentative experimentation", while warning that it was "likely to be indicative of its future exploitation."^[7]
• Governance matters more than technology. Feldstein's data show that advanced democracies use AI surveillance at a higher rate than closed autocracies. He places the deciding factor in the quality of governance rather than in the technology.^[4]
• Democratic uses. A 2026 Carnegie report by Micah Weinberg described how AI tools can help large-scale public deliberation. Examples include Polis, an open-source opinion-mapping tool used in Taiwan's vTaiwan process, and the processing of more than 100,000 public responses for the 2023 UK AI Safety Summit.^[8] Weinberg also warned that AI systems applied to social interaction "are generally designed to maximize engagement, which tends to boost emotional and divisive content," and that democratic institutions lack the capacity to deploy such tools responsibly.^[8]

Open questions

• Effect on political outcomes. The reports above document far-right production and circulation of AI content, but not whether it changes votes, recruitment, or violence compared with content made without AI.^[7][3] Studies comparing audience responses to AI-generated and conventional propaganda would help answer this.
• Whether AI is special. McQuillan and Watkins argue that features of AI itself favor fascist politics. Their critics see continuity with older critiques of bureaucracy or technology.^[1][6] Comparing how AI and earlier technologies (radio, film, databases) were adopted by authoritarian movements could clarify what, if anything, is new.
• Influence of investors' ideas. Thiel has written that freedom and democracy are incompatible and has funded people associated with neoreaction, and Andreessen has named practices such as "trust and safety" and "tech ethics" as enemies.^[12][2][13] How much these ideas shape the design of AI products, company policies, or government contracts is less clear. Answering this would require evidence from company decisions and contracts, not only from public writings.
• Current surveillance data. Feldstein's index dates from 2019, before the spread of generative AI, and needs updating.^[4]

Analysis: effects on human agency

This section contains value judgments based on the evidence cited above.

The evidence supports concern on several fronts, without requiring a conclusion that AI is inherently fascist.

AI propaganda and fake personas reduce agency by manipulating what people believe about their situation. ISD's finding that reported far-right AI content was neither removed nor labeled a month later means audiences got no signal from the platforms about what was synthetic.^[3] Content promoting forced deportation of minorities targets the most basic agency of the people it names.

AI surveillance concentrates the ability to watch and predict behavior in the hands of states and suppliers. Feldstein's argument that governance quality determines whether it becomes repressive implies that legal limits, independent oversight, and transparency are what protect agency here.^[4]

Ownership and control of AI systems also matter. When investors who have written that freedom and democracy are incompatible, or who describe safety practices as "the enemy" and slower AI development as a cost in lives, fund and shape AI companies, their views can influence tools that governments use on the public — such as ICE's use of Palantir software to prioritize people for deportation — while the people affected have little say in how those tools are designed.^[12][13][16] The Grok incident showed that a single company's configuration change could put extremist content in front of a platform's users for 16 hours.^[19] Weinberg's account of AI-assisted deliberation shows that the same class of tools can be used to widen participation, when institutions have the capacity and standards to govern them.^[8]