Comment from William Bowles:
On your article: “AI researcher warns machines could take power. The Pentagon already uses AI to choose targets”
Yeah but it’s people, corporations and their servants, not the machines that kill. You have a responsibility to make it clear that AI is NOT intelligence, they’re just machines loaded with stolen human skills, not perpetuate the myth of machines being intelligent and “taking over” from humans. This myth is being deliberately fostered in order to hide capitalism’s role in the process.
William Bowles
ed****@***********es.info
Response from Gary Wilson:
William Bowles is right, and he is right about my headline. No machine decided to bomb Iran. No machine decided to put Palestinians on a target list or to track workers through their phones. Corporations build these systems and sell them. Washington buys them and uses them. My headline repeated the warning that machines could take power. That warning serves the owners of the machines, and putting it at the top of the page helped it along.
Bowles asks for more than agreement, and he should have it. He says what is called artificial intelligence is not intelligence, just machines loaded with stolen human skills. The second half of that sentence is the half that matters, and Marx gave us better words for it.
These systems are made out of human labor — the writing, translating, drawing, photographing, coding and talking of many millions of people, taken without payment and without asking. Marx called machinery dead labor: past work that meets living workers as somebody else’s property, and that lives, vampire-like, by drawing on living labor. In the Grundrisse he described how the knowledge built up by all of society gets put into the machines, and how it turns into private property along with them. The monopolies have done exactly that. They have taken what humanity has learned over thousands of years and now rent it back to us by the month.
So the argument does not turn on whether the machines think. Whatever they can be said to do, they did not come by any of it on their own. It is human work, congealed and privately owned. The name “artificial intelligence” hides whose intelligence it is. Ask who owns it and what it is pointed at, and the mystery drains out of it fast.
What it is pointed at is not only Iran.
Maven — Palantir’s Maven Smart System — sorted target lists when Washington opened the war on Iran on Feb. 28, 2026. The same company built a tool for Immigration and Customs Enforcement called ELITE, for Enhanced Leads Identification and Targeting for Enforcement. It maps the people to be picked up, assembles a dossier on each one and scores how likely the person is to be found at a given address. It draws on Medicaid records; ICE gained access to the health data of some 80 million people enrolled in the program. Palantir is also building a platform called ImmigrationOS that pulls immigration files, travel records and license plate readers onto one screen.
The target list for the bombing of Iran and the raid list for Minneapolis come out of the same shop, built on the same principle: gather everything known about a person, rank it, hand a commander the list.
It does not stop at the war command and the deportation squads. Health insurers run claim denials through prediction software. Landlords screen tenants with it and employers screen job applicants. The ride and delivery platforms use it to set what a driver gets paid for a given trip. In the warehouses it sets the pace of the work and flags the workers who fall behind. Meta and Microsoft have cut tens of thousands of jobs while pouring money into computing power at the same time. Technology built out of the labor of millions is being used to speed them up, hold their wages down and throw them out.
One episode shows where the decisions actually get made. Claude, made by Anthropic, was built into Maven in late 2024 and was running on classified Pentagon networks during the Iran strikes. Anthropic tried to hold two limits: no mass surveillance of people inside the U.S., and no fully autonomous weapons that pick and fire on targets with no human decision. Washington would not have it. On Feb. 27, 2026, Trump ordered federal agencies to stop using the company’s technology. In early March the Pentagon designated Anthropic a “supply chain risk,” a label normally kept for firms tied to foreign adversaries, and began testing replacements from OpenAI, Google and Musk’s xAI. Hegseth had already said in public that the Pentagon would not use AI models that will not let you fight wars. Anthropic sued, and on Aug. 27, 2026, a federal judge ordered the designation lifted as unlawful retaliation.
Take the measure of that. A corporation asked for the mildest imaginable limit on how its product could be used in war, and the government treated it as an enemy for asking. Then another arm of the same government handed the company its victory. The bombing went on through all of it, and Maven and Claude went on sorting the target lists. The fight was over the terms of a contract, and neither side ever raised a question about the war. Amodei said as much in the statement he issued on Feb. 26, 2026, the day before Trump’s order: Anthropic has “never raised objections to particular military operations.”
The people who build these systems are not all willing. On May 5, 2026, workers at Google DeepMind in London wrote to the head of Google’s British operations demanding recognition of the Communication Workers Union and Unite as their representatives. Nearly all the union members there backed it. What they want is to stop Google’s AI from being handed to the U.S. military and to the Israeli forces that Washington arms and directs, and they have raised the prospect of research strikes — refusing to work on the company’s main products. It would be the first union anywhere in the world at one of the corporations building these systems.
The ruling class gains from presenting all of this as a force of nature standing above society. It hides the owners and the class interests giving the orders — the Wizard of Oz trick: “Pay no attention to that man behind the curtain.” The myth runs in two directions at once. Investors are promised machines that will soon do almost anything. The public is warned that those same machines may slip out of human control. Both stories raise the price of the machines and the authority of the corporations that hold them. Both keep the curtain closed.
Jacob Coxon, the researcher who resigned from Anthropic on Sept. 8, 2026, warns that the machines may take power. The owners of the machines have power now and are using it. Maven sorted the targets for the bombing of Iran. ELITE sorts the addresses for the raids here. Neither one chose anything. The men who chose held a press conference about it on March 10, 2026. And the workers in London refusing to build the next one are making a choice too — the only kind that has ever stopped anything.
Gary Wilson
Join the Struggle-La Lucha Telegram channel