A Letter to the Left on AI

The Pattern

To the left,

I am writing this because I keep watching people I agree with lose a fight they do not realize they are in.

The left has a position on AI. The position is suspicion. The suspicion is correct. Pew found 40% of Americans expect AI to harm society over the next twenty years. The suspicion is also not a strategy.

What the suspicion gets right

The companies building the most powerful AI systems are not neutral actors. OpenAI, Anthropic, Google, Meta, Palantir. They are firms with shareholders, lobbying budgets, and a structural interest in framing their product as inevitable. The frame is not accidental. It is the most expensive public relations campaign in the history of computing, and it is working. The FTC has already had to launch Operation AI Comply, a crackdown on companies making fraudulent AI claims, including one firm selling “AI” that was resold email lists. The industry lies about what the technology does. That is documented. It does not follow that the technology does nothing.

The labor threat is documented, not speculative. Anthropic’s own CEO told Axios in May 2025 that AI could wipe out half of entry-level white-collar jobs and push unemployment to 10-20% within five years. He later walked back the specifics. The direction of the warning did not change. A working paper tracking corporate filings found AI named as a factor in roughly 76,000 position eliminations in 2025 alone, and that figure is almost certainly undercounted because firms have learned not to say the quiet part in SEC disclosures. Challenger, Gray & Christmas tracked 55,000 cuts explicitly attributed to AI in the same year, a more conservative count that only counts what companies state on the record. Goldman Sachs’s workforce analysis estimates hundreds of millions of jobs exposed globally. The exposure is not evenly distributed. It falls first on the clerical, the paralegal, the copywriter, the junior analyst, the translator, the customer service worker. The jobs that were supposed to be the ladder.

The environmental cost is also documented. A PBS/AP investigation found AI data centers now rival entire nations in energy and water consumption. The water is not abstract. It is drawn from aquifers in counties that did not vote for this and were not asked.

The surveillance dimension is the one the left should be loudest about and is somehow the quietest on. Palantir, the company that built the Investigative Case Management system ICE used to build profiles of immigrant families for deportation during Trump’s first term, is now an AI company. The same firm that told The New York Times it did not work with ICE’s Enforcement and Removal Operations, and was then caught doing exactly that by documents obtained through FOIA, is selling predictive policing and immigration targeting as a product line. The product is the pattern. The pattern is the threat.

So the suspicion is not paranoia. It is the correct reading of who owns the machinery and what the machinery is for.

What the suspicion does not get you

Here is the problem. Suspicion is not a program. Suspicion does not write a bill. Suspicion does not organize a workplace. Suspicion does not build the tool a tenant organizer uses to scan a thousand leases in an afternoon for illegal clauses, which a tenant organizer I know is now doing with a model she runs on a laptop. Suspicion does not train the model a public defender’s office uses to draft motions faster so the office can take more cases. Suspicion does not produce the translation layer that lets a workers’ center serve members in seven languages without a six-figure contract.

The largest field study on AI in the workplace, the NBER paper by Brynjolfsson, Li, and Raymond, found AI assistance raised customer-support productivity 14% on average and 34% for novice and low-skilled workers. The gains flowed to the bottom of the skill distribution. A Harvard randomized trial found students with a tailored AI tutor showed roughly double the learning gains of a traditional classroom. These are not marketing claims. They are measured results. The left should be the faction that knows the difference and uses it.

The right is not having this crisis. The right is building. The right has a $140 million super PAC funded by executives affiliated with OpenAI and Palantir paying influencers to stoke China fears and drive adoption. The right has the libertarian tech press treating regulatory caution as weakness and acceleration as seriousness. The right has Marc Andreessen telling anyone who will listen that the only sin is to slow down. The right has a coherent, funded, loud position. The position is: build it, own it, deploy it, and let the market sort the bodies.

The left’s position, as near as I can tell from the discourse, is: do not touch it, and if you do touch it you are contaminated.

That is not a position. That is a purity test. Purity tests do not win infrastructure fights. The side that builds the infrastructure writes the rules. The side that refuses to engage spends the next decade filing comments with an agency the other side captured.

The Hank Green problem

I watched what happened to Hank Green in August 2026. He used ChatGPT to find academic papers for a video. He did not use it to write. He did not publish AI text as his own. He used it the way a researcher uses a search engine, and he disclosed it. The response from parts of the left was to treat him as a traitor to human creativity. He issued an apology. The Verge reported his channel might pause. Kotaku ran a headline calling his AI use “unhealthy.”

The man used a research tool and the left ate him.

This is the failure mode. The left has built a culture where any contact with AI is treated as moral compromise, and the result is that the people most likely to use the tools ethically, the ones who would disclose, who would cite, who would refuse the grift, get driven out first. The people who do not care about disclosure remain. The tools do not get less powerful because good people refuse to touch them. The tools get used by worse people with less oversight.

The purity frame is a self-disarming mechanism. It functions as abdication wearing the costume of principle.

The actual question

The question is not whether AI should exist. It exists. The question is not whether AI will be deployed. It is being deployed. The question is who owns it, who profits from it, who is displaced by it, who is surveilled by it, and who writes the rules that determine all of the above.

Those are distributional questions. They are class questions. They are the questions the left exists to answer. And the left is answering them with a gesture of refusal that concedes every one of them to the firms and the financiers.

Bernie Sanders introduced the American AI Sovereign Wealth Fund Act, a proposal to socialize the gains of AI rather than hand them to shareholders. I have seen left discourse treat it as unserious, or as too late, or as not pure enough because it does not call for abolition. A proposal to own the means of generation collectively, and the response from the left’s own base is to litigate whether the proposal is sufficiently clean. This is the pattern. The perfect kills the possible. The possible was the only thing that was going to help anyone.

What I am not saying

I am not saying the danger is overstated. The danger is real. The labor displacement is real. The environmental cost is real. The surveillance application is real and getting worse. The concentration of compute in three or four firms is a power arrangement that should alarm anyone who claims to care about monopoly.

I am not saying every left critic of AI is a purity-tester. Some of the sharpest critics, the labor scholars tracking displacement, the data center organizers fighting water extraction in their counties, the privacy lawyers fighting Palantir contracts, are doing the actual work. They are not the target of this letter. They are the model for it.

I am not saying the left should become accelerationist. Accelerationism is the right’s frame, and it is a frame designed to make refusal feel naive and engagement feel sophisticated. The left does not need to adopt the opponent’s frame to enter the fight.

What I am saying

The left is losing a structural fight by refusing to show up. The refusal is dressed as resistance. The consequence of the refusal is that the right builds the infrastructure, the firms write the policy, the courts defer to the engineers, and the people the left claims to represent lose their jobs to systems no one on the left bothered to shape.

There is a version of left AI politics that is serious. It demands worker control over deployment decisions. It demands public compute. It demands a ban on predictive policing and immigration targeting, not a ban on the underlying math. It demands that displaced workers own a share of the productivity gains that displaced them. It demands data center siting be subject to local consent and water-impact review with teeth. It demands that the firms training on copyrighted work pay into a public fund, not settle with individual publishers and call the matter closed. It demands privacy law that reaches training-data extraction at ingestion, not disclosure-and-consent theater after the fact. It demands retroactive encryption of the surveillance archives that already exist — Canary Mission’s database, Flock’s movement records, Palantir’s deportation profiles — because the lever is not delete and the archives are not going to be deleted. It demands encrypted inference as the baseline for any AI tool used by organizers, public defenders, and workers’ centers, because using the tool on terms that feed the surveillance dividend is not engagement. It is surrender with a cleaner conscience. The full policy framework is in Privacy Law in AI Is Infrastructure Policy.

That politics requires engagement. You cannot regulate what you do not understand. You cannot demand worker control of a deployment process you have never been inside. You cannot write a bill about compute concentration if your movement treats anyone who reads a model card as suspect. Demand for entry-level AI skills has nearly tripled since fall 2025. The jobs are changing whether the left shows up or not. The question is whether the people who lose work to these systems get a share of what the systems produce, or whether they get nothing and a lecture about why they should have been more careful.

The tools are not going to wait for the left to feel clean about using them. The tools are already being used. The only question left is whether the left is in the room when the decisions get made, or outside the room holding a sign that says the room should not exist.

The room exists. The decisions are being made. The sign is not stopping them.

Globally, the Stanford AI Index found 59% of people say AI’s benefits outweigh its drawbacks, and the number is rising. In developing countries the figure is above 75%. The countries with the least access to healthcare, education, and legal services look at the same technology the American left treats as contaminated and see the first chance in a generation to close the gap. They are not wrong. The left that refuses to engage is not just ceding the fight to the right. It is ceding the future to everyone who decided the tool was worth using while the left was still deciding whether touching it was clean.

Sources and scope

This letter draws on polling, labor research, and documented corporate conduct current as of August 2026. The Sanders proposal and the cited displacement research are specific legislative and academic artifacts, not gestures. The Hank Green episode is referenced as a case study in movement culture, not as a personal judgment of any individual participant in that discourse.