AI as a Fascist Artifact

(This is a bit of a merger of two talks I recently gave about fascism and AI. One was in German at the Cables Of Resistance conference, one in English at the Milton Wolf Seminar on Media and Diplomacy. I added some shots of the slides I used as a structure for the text which might make it look a bit weird. You can just ignore the images if you want to. They are kinda like subchapter marks. The text is not exactly what I said but a longer version of my arguments that should be easier to read.)Our world and our access to it is increasingly structured through technological mediation: Digital platforms and systems are a massively important aspect of not just our work environments or our interactions with government entities or “the media” but also our individual interactions with one another. Our world is built around technological infrastructures that define what we see, who we can talk to and what information gets presented to us.We also live in a time of growing fascist threats all over the planet: Many countries have neofascist movements and parties trying to gain power and potentially even get conservative parties to include them in governments. Some even have had success. Fascism is back with a vengeance. (Antifascists have been warning for decades but that realization sadly doesn’t help anyone. Maybe after we’ve gotten rid of the fascists we can learn something from that.)And of course we are living in the “AI” age, where stochastic systems with attributed agency are being pushed – currently under the moniker of “agentic AI” – into all our professional and personal workflows. “AI” is the singular focus of the tech sector currently and the magic technology that governments and companies are putting all their hopes into for figuring out how basically keep late stage capitalism going on for a bit longer.In this text I want to analyze the relationship of fascism and what is called “AI” these days. Is this “technology” that keeps being used to reshape the world around us (for better or worse but dominantly worse) in some way connected to fascism? Or is it just something fascists like to use? Is it neutral?When we think about fascism we often do that by looking at the actors: Evil individuals doing evil things for evil reasons.

And that is often how we are looking at the relationship of “AI” and fascism: We see Trump’s White House and other parts of his administration using generative “AI” to create openly fascist propaganda about their leader, using “AI” to manipulate photos to make their opposition look bad and using “AI” to in generally increase the amount of racist and fascist media in the world.Palantir’s CEO Alex Karp has for about 2 years been unable to talk about anything but how he wants to use his data integration platform (Palantir’s product is quite boring TBH) to kill people. And not just random people. He frames himself and his software engineers as warriors for “the West” who are protecting the USA and “the West” against “the enemy”. Palantir openly wants to be a critical part of the military’s infrastructure that makes “kill decisions”, wants their software to be treated and seen like a weapon – which aside from being a very fascist appeal to the normalization of violence – also can be read as a sales pitch trying to bolster the software’s capabilities and power: Nobody will spend billions on simple data integration. But if it can “kill your enemies” maybe the contracts will keep rolling in?And finally we have people like Marc Andreessen who last year published a “Techno-Optimist Manifesto” which – in contrast to what the title points at – is mostly a document based on his demand to not be regulated or laughed at by people smarter than him. But it’s not just the somewhat reductionist views on what “AI” and other technologies he is invested in can do and will do for the world, the document is remarkable because it directly and openly quotes and bases its reasoning on the writings of Italian fascists and other right wing reactionaries such as Nick Land and because it explicitly marks “the enemies”: The communists, the luddites and those who want to regulate tech. Basically the go-to enemies of fascism since its inception.This realization of a “capture” of tech or specific technologies (like “AI”) by the right sometimes leads to people wanting to “save” or “take back” those technologies. Because they are so deeply embedded into our lives, because we’ve gotten so used to them that conceptualizing a life without them seems impossible.

We like our apps and convenience. And it’s not those apps’ and technologies’ fault that fascists keep using them. Maybe if the left stopped criticizing “AI” (or the Metaverse or Blockchain or whatever) then we could make “AI” good and ethical and democratic? Maybe we can save those technologies from the bad people? Lead it back into the light? Maybe if we made it Open Source?In his influential 1980’s paper “Do Artifacts Have Politics?” Langdon Winner argues that this view of “neutral technology” does not hold up. That the politics of specific artifacts do not just come from who uses the technology and for what purpose but that technologies have built-in politics that stem from the political views and goals of the people building the technology as well as their internal structure.He shows this by pointing at how certain bridges were built racist: When the civil rights movement in the US got black kids the right to go to the often better schools that used to only accept white kids, politicians did for example plan roads and bridges in a way that the buses that were supposed to take the black kids to the white schools could not pass the bridges and roads. This was not oversight but design intent. The racism is built into the structure of the artifact itself.Winner also argues that certain technologies imply certain political or social structures in order to exist: The nuclear bomb implies not just scientists who can build it and a state thinking that that form of destruction is a valid form of acting in the world but also a security state capable of controlling and defending it. You simply cannot build a nuclear bomb without those structures, they are implied if not required, enforced by the artifact itself.Winner’s work does not argue that the embedded politics of an artifact are always absolute: We do know of many potentially oppressive technologies that have been taken by artists and activists to turn them against their original use. But that is always an uphill battle: Surveillance will always lean towards a more forceful, rigid, less free understanding of government for example. You can use (counter-)surveillance of course but you always have to be aware of not reproducing the logic you are trying to criticize or attack.Drawing from Winner’s insight the question emerges, what the embedded, structural politics of “AI” are? What world, what view of the world, what politics does “AI” require or imply?

What’s the path that “AI” as we understand it today put us on?Before we dive into this I want to quickly talk about the definition of the term “AI”. I do not think that “AI” is a very useful term – TBH I would mostly advise against using it in general because it clouds more than it explains or makes clear. But still the term is everywhere so we have to deal with it. And one important realization about “AI” is that it’s not a very well-defined term: “AI” can be an LLM (a stochastic token extruder), a system of symbolic knowledge representation, an Excel macro, a person in a call center in India or just a slide in a pitch deck. “AI” doesn’t mean anything specific. At least not a specific type or class of technical artifacts.I am a big fan of Ali Alkhatib’s definition of AI: I think we should shed the idea that AI is a technological artifact with political features and recognize it as a political artifact through and through. AI is an ideological project to shift authority and autonomy away from individuals, towards centralized structures of power. Projects that claim to “democratize” AI routinely conflate “democratization” with “commodification”. Ali Alkhatib“AI” is a political project – I have also sometimes called it a narrative – whose purpose is the shifting of power and agency away from people and organizations towards centralized power structures. These centralized power structures are currently mostly a handful of big tech corporations and the “AI Labs” they keep shoveling money into.So while I don’t think that “AI” is a great term to use, we will keep using it for the rest of this text in the understanding that dominates the term right now: In that reading “AI” stands for a class of stochastic machine learning systems that can store and apply patterns extracted from data in order to do either pattern recognition (think computer vision) or (and that is the dominant narrative vehicle today) as generative systems (“generative AI” or “genAI”). So when I write “AI” think ChatGPT or Claude or Gemini or Deepseek, etc.So, back to fascism.There is of course a huge load of research and analysis from media studies and related fields about the fascist use of “AI”: I specifically recommend Gareth Watkins’ essay “AI: The New Aesthetics of Fascism“.

In it, Watkins shows that there are properties in the structure of the output from generative image extruders that align well with the politics and reasoning of the right.“AI” is built by scraping the Internet and any other data source one can find and most of that data is heavily racialized, is based on a colonial, sexist, heteronormative understanding of the world and the past. There literally is no police data that’s not racist. If you base your image generator on the images available, LGBTQIA representation, representation of people not conforming to the social expectations of acceptability is lackluster at best for example.And all that data does by definition exclude: “AI” is not built on “all of humankind’s knowledge” but based on whatever a mostly western view of the world and what is relevant looks like. Cultures who are not within that framework, who might even be based on more oral forms of keeping history and knowledge are not represented. Even if those groups are not actively excluded (which again they very often are) there are huge populations who just are not seen by the data and do not get a say in how they are represented. Or if they are represented it’s just as problems: Think about unsheltered people for example.The right loves those patterns because they confirm their prejudices: Ask an image generator for a picture of two people kissing and you most often get a heterosexual couple, often white. Because that’s what the training data looks like. That makes “AI” perfect for creating the form of idealized, fictional “past” that fascists love to allude to (“make America great again“), a past that never existed but that needs to be saved or restored (we’ll get back to that later).But there is another aspect of “AI” usage that fuels the right’s enjoyment of using “AI”: It hurts the people they want to hurt. “AI” is currently mostly used to generate media (think images, illustrations, music or text). But traditionally people in those creative industries are more left leaning, more inclusive. Fascists just can’t create good and interesting art. Using “AI” to take that groups’ jobs, their livelihood, their creative expression away is exactly why using “AI” to create an image is so enjoyable by right wingers: It’s a vulgar display of power.And this perfectly leads us to looking at the structural properties of “AI”.