In April, a 20-year-old from Spring, Texas, firebombed the home of OpenAI CEO Sam Altman and was arrested outside the company’s headquarters carrying kerosene and a manifesto warning of humanity’s extinction, along with a list of AI executives’ names and addresses. Days earlier, someone fired thirteen rounds into the home of an Indianapolis city-county councilman who had backed a data center, leaving a “No Data Centers” note behind. Violence connected to artificial intelligence has arrived, and it is worth being clear-eyed about what it is and where it is likely to go.
The useful starting point is that none of this is unprecedented. Technological displacement has produced political violence before, most dramatically during the Industrial Revolution, and the way it unfolded then is a reasonable guide to what to watch for now. When speakers reassure audiences that AI is just the next industrial revolution, they are reaching for comfort, but the Industrial Revolution was one of the most violent economic transitions in modern history. It raised living standards over the long run, and along the way it produced machine-breaking riots, decades of labor unrest, and the movements that carried that grievance into revolutionary politics. The analogy is apt. It is just not reassuring.
That history offers three practical lessons for reading the current moment.
The first is that displacement violence comes in two forms on two very different clocks. In the short term it is local, immediate, and disorganized: a specific target, a specific grievance, a bounded act. The Luddites breaking the specific machines that undercut their trade are the classic case, and the Indianapolis shooting fits the same shape, a single official, a single project, a single demand. This kind of incident is the one to expect first, and current threat reporting confirms it is already happening. A recent West Point Combating Terrorism Center study found that anti-AI grievances are shifting from distant executives toward local officials, because a data center turns an abstraction into a facility next door that consumes a specific neighborhood’s water and power. In the long term, the more serious threat is different: dispersed grievance captured by an organized movement that carries it for years and reframes it as a fight against the whole system. The anarchist wave of the late nineteenth century, which assassinated six heads of state, and the communist movement grew out of exactly that slow process.
The second lesson is counterintuitive: organization usually prevents violence rather than causing it, at least at first. A grievance with an organization behind it has somewhere to go—collective bargaining, lobbying, litigation, the vote, and leaders with a stake in keeping it legal. New York’s taxi and rideshare drivers, facing displacement by robotaxis, did not attack anything. Through their 28,000-member union they pushed until the state pulled back a robotaxi proposal and the city let a testing permit lapse. A negotiated outcome is a healthy outcome, and it is the common one when a channel exists and delivers. The danger is grievance with no channel. The blue-collar trades eventually built unions; the white-collar and newly-graduated knowledge workers now watching AI take entry-level jobs have almost none. Right now that keeps them quiet, booing AI boosters at graduation rather than organizing. But an unorganized, unresolved grievance is precisely the raw material that becomes dangerous the moment someone organizes it around the claim that the system will never help. That is how riots became movements the first time.
The third lesson concerns the rarest and most dangerous actor. Most displacement violence is driven by personal grievance, someone losing a job, wages, a neighborhood. That kind of anger has a natural ceiling, because there is a concrete remedy the person actually wants. But a small number of actors are driven by something with no ceiling: the belief that the technology itself threatens humanity. The Altman attacker had no personal stake anyone has identified. Theodore Kaczynski, the Unabomber, had none either, and sustained a campaign for seventeen years. When someone sincerely believes a technology will end the species, almost any act can seem justified, and the ordinary limits fall away. This kind of violence is uncommon, because few people hold that belief intensely enough to kill for it, but it echoes a pattern documented across apocalyptic and millenarian movements, where conviction that catastrophe is imminent has repeatedly driven people to violence they would otherwise never consider. That makes it severe and hard to deter when it comes. What is new about this moment is that the framing is no longer confined to lone attackers. AI’s own leaders now describe their technology in terms of mass unemployment and human extinction, putting that catastrophic framing into wide circulation.
These are not distant concerns. The Soufan Center warned in November that online threats to sabotage data centers were multiplying across economic, environmental, and ethical motivations, and reported in May that the rhetoric was turning toward direct threats against the officials who approve projects. The Indianapolis note is a good example of how little a single incident tells us on its own. Two words do not reveal whether the shooter was angry about his own utility bills, which is a contained grievance that ends if the project stops, or acting on a broader conviction that does not end when one facility is stopped. At the moment of the act, the motivation is often invisible, and the motivation is what determines how dangerous the person is.
The practical problem is that no single office is set up to see the whole picture. Local police handle a shooting like the one in Indianapolis as an isolated crime. Federal prosecutors decide how to charge an attack like the one on Altman. Task forces built for organized extremist movements are not looking at lone actors or at unorganized displaced workers. The analysts watching online anti-AI sentiment, the lawyers handling the training-data lawsuits, and the investigators reading manifestos rarely compare notes, and no one is treating the largest displaced group, this year’s graduates, as a security question at all. The answer is not a new agency. It is a shared awareness that these scattered incidents are pieces of one familiar pattern, that the immediate risk is localized violence around specific projects, and that the longer-term risk is what happens if today’s disorganized grievance finds organization and a cause. The Industrial Revolution showed how this goes. The value of knowing that is the chance to watch for it early.
