On April 10, 2026, a 20-year-old man from Texas drove to San Francisco, threw a Molotov cocktail at Sam Altman’s home, set an exterior gate on fire, and then walked three miles to OpenAI’s headquarters and threatened to burn that down too. Two days later, a second incident at the same address ended in two more arrests. The suspect in the first attack, Daniel Alejandro Moreno-Gama, was charged with two counts of attempted murder, attempted arson, and possession of an incendiary device. He was found in possession of a manifesto he had written opposing artificial intelligence and listing the names and addresses of other AI executives, according to the federal criminal complaint. His defence attorneys said he was in the middle of an acute mental health crisis.
These are the facts. They are not ambiguous. The acts described are criminal, dangerous, and indefensible. Nothing in this analysis should be read as suggesting otherwise.
But reporting only the criminal facts and stopping there misses the story that actually matters, which is the environment those acts occurred in. That environment is documented. It is not speculative. And as we covered in our analysis of the Stanford 2026 AI Index, the data describing it has been sitting in plain sight for months.
What the Numbers Actually Show About Public Sentiment
The Stanford 2026 AI Index included a finding that received far less attention than it deserved: 73% of US AI experts view AI’s impact positively. Only 23% of the general public shares that view. That is not a disagreement about facts. It is a 50-point gap that reflects two genuinely different experiences of the same technology.
AI experts sit inside the 20% of companies and institutions where AI is generating returns, as documented by PwC’s 2026 AI Performance Study. Their salaries have increased. Their career prospects have improved. Their tools work. The technology is delivering for them personally, and they report that honestly when surveyed.
The 77% of the public that does not view AI positively is also reporting honestly. Their entry-level colleagues are disappearing. Their employers are cutting headcount and citing AI productivity as the justification. Their applications are being filtered by algorithms before a human reads them. The gains they were told AI would deliver have not arrived at their pay grade. They are experiencing the disruption side of the ledger without access to the windfall side.
The Industrial Revolution Parallel That Fortune Got Right
Fortune’s coverage drew a parallel to the Industrial Revolution and the Luddite movement of the early 19th century. It is worth taking seriously rather than using as a dismissal, which is how the comparison usually functions in tech industry commentary.
The Luddites were not irrational opponents of technology. They were skilled textile workers whose livelihoods were being destroyed by machinery owned by others, with no redistribution of the gains and no transition support from the state. Their violence was wrong. Their underlying grievance was entirely rational and was later validated by decades of economic evidence about wage suppression and working conditions during industrialisation. History has been considerably kinder to their analysis than to the framing used to suppress them at the time.
The parallel is not precise. AI’s disruption is faster, broader, and more concentrated in white-collar work than industrial machinery was. But the structural dynamic is similar: productivity gains captured by capital, disruption borne by labour, and a growing gap between the people inside the system of beneficiaries and the people outside it.
The Specific Economic Data Behind the Frustration
This is not abstract. CNBC confirmed that Moreno-Gama’s written document advocated against AI and its executives specifically. His stated grievance was about humanity’s extinction from AI development. That framing is apocalyptic and disconnected from what the evidence actually shows.
But the economic evidence that generates frustration without needing apocalyptic framing is substantial. Goldman Sachs documented that AI is eliminating roughly 25,000 US jobs per month while adding back 9,000, for a net monthly loss of 16,000 positions. As we covered in our analysis of Goldman’s scarring data, workers displaced by technology suffer cumulative earnings losses nearly 10 percentage points below never-displaced workers over the following decade. Jack Dorsey cut 40% of Block’s workforce in a single announcement and his stock surged 24%. As we covered in our analysis of the Block layoffs, the financial incentive structure now rewards these decisions regardless of whether the AI tools actually replace the workers being cut.
None of this justifies violence against any individual. All of it explains why the public’s trust in AI and in the people leading AI companies is at 23% and falling.
The Response That Makes the Problem Worse
The dominant response from the AI industry to rising public hostility has been to frame it as ignorance, fear, or mental illness. When a survey shows 77% of the public views AI negatively, the explanation offered is that people do not understand the technology. When someone acts violently, the explanation offered is mental health crisis, which in Moreno-Gama’s case may well be true and is legally relevant.
The problem is that this framing removes any obligation to address the documented economic reality that the 77% are responding to. If public hostility is purely a communication failure or a pathology, the solution is better messaging, not different outcomes. And if the problem is messaging, the people doing the messaging do not need to change anything about who captures the gains or who bears the disruption.
The Washington Post’s coverage described the attacks as a symptom of a growing division around AI. That framing is more accurate than the mental health framing alone. Divisions do not close by dismissing one side as uninformed. They close when the underlying distribution of costs and benefits shifts enough that the people bearing the costs can see a credible reason to trust the outcome.
That shift has not happened. The Stanford data and the Goldman data both show it has not happened. The attacks on Sam Altman’s home are a data point in a distribution of public responses to AI that runs from scepticism to organised opposition to, at the extreme and irrational end, violence. The extreme end should be reported and prosecuted. The rest of the distribution should be taken seriously as a rational response to documented economic outcomes rather than explained away as something that better PR will fix.
This article draws on reporting by NPR, CNBC, The Washington Post, and Fortune. Analysis and interpretation reflect the author’s reading of publicly available information. Featured image: Sam Altman speaking at TED. Photo by Steve Jurvetson, CC BY 2.0, via Wikimedia Commons.