Andrew Yang is back. His pitch: tax AI, not labor. On CNBC’s Power Lunch, he argued the government should shift its revenue base from payroll to artificial intelligence. The logic is seductive. The data is missing. The execution is a structural impossibility.
Yang built his political brand on automation warnings during his 2020 campaign. He proposed a universal basic income called the Freedom Dividend. He backed cryptocurrency adoption and clearer digital asset rules. Now, as CEO of Noble Mobile and co-founder of the Forward Party, he renews the call. He points to Anthropic CEO Dario Amodei’s 3% AI revenue tax idea from 2025. Amodei said the levy would apply each time a model generates revenue. Yang says the same logic should apply broadly. Firms would weigh AI costs against payroll costs. They would choose humans.
The data paints a grim picture. A CNBC and Generation Lab survey from August 13 polled Americans aged 18 to 34. 45% expect AI to hurt their careers. Only 10% expect it to help. Bridgewater Associates executives Greg Jensen and Nir Bar Dea wrote a New York Times opinion piece. They estimated AI could displace 18% of current US jobs within five years. They used that estimate to back their own AI token tax proposal. Customer service, which employs roughly 2.9 million Americans, is already shedding roles. The shift is visible. Yang proposes sending the tax revenue directly to workers as checks. He says retraining programs rarely work. He points to failed efforts for coal miners and warehouse staff.
I have spent years auditing smart contracts. I have seen how code hides intent. I have seen how revenue can be obfuscated across chains. The idea of taxing AI revenue is a political fantasy. It assumes a centralized registry of all AI-generated income. That registry does not exist. It cannot exist without breaking the open nature of the internet. AI models are embedded in supply chains. A model generates revenue when a customer buys a product. The model’s contribution is a fraction. How do you tax the fraction? You cannot. You end up taxing the entire transaction. That is a sales tax, not an AI tax.
The proponents ignore the enforcement problem. Yang’s logic is clean: tax the machine, not the worker. But the machine does not have a wallet. It does not have a tax ID. The company that owns the machine does. So the tax is on the company’s revenue. That is a corporate income tax with a different label. The only difference is the rate. Yang wants to lower payroll taxes and raise AI taxes. That is a tax shift, not a new instrument. The structural question is: can you distinguish between revenue generated by AI and revenue generated by human labor? In a modern company, the lines are blurred. A software engineer uses AI tools. A call center uses AI scripts. The tax base becomes a matter of accounting discretion. Auditors will fight over definitions. The IRS will lose.
I have seen this pattern before. In crypto, we tax transactions. But we tax them at the point of sale, not at the point of generation. The AI tax proposal inverts that. It attempts to tax the input, not the output. That is like taxing the electricity used by a GPU instead of the crypto it mines. The GPU miner can switch to a different chain. The AI company can shift revenue to a subsidiary. The tax becomes a game of regulatory arbitrage. Hype burns hot; logic survives the cold burn.
What the bulls got right. Fear of job displacement is real. Bridgewater’s 18% estimate is plausible. The current payroll tax system is a disincentive to hire humans. That is a design flaw. But the solution is not to tax AI. The solution is to fix the payroll tax. Reduce the burden on employers. Fund social programs through a broad-based consumption tax or a data dividend. Yang’s crypto advocacy suggests he understands blockchain. A smart contract-based UBI distribution could work. But that is a separate mechanism. It does not require an AI tax. It requires a funding source. Taxing AI revenue is a poor funding source because it is unenforceable.
I do not fix bugs; I reveal the truth you hid. The truth is that Yang’s proposal is a political signal, not a policy. It signals a willingness to confront automation. It signals a desire to protect workers. But it hides the complexity of implementation. Every gas leak is a story of human greed. Every tax proposal is a story of human avoidance. The AI tax will be avoided. Companies will restructure. They will classify expenses. They will move revenue offshore. The tax will collect less than projected. The checks will be smaller than promised. The displaced workers will still be displaced.
The structural impossibility is clear. You cannot tax what you cannot see. AI models are opaque. Their revenue contributions are indeterminate. The tax base is a moving target. The only way to enforce it is to require all AI models to be registered and audited. That is a centralized database. That is a registry of all AI transactions. That is a surveillance system. It is the opposite of the decentralized ethos Yang once championed. The irony is thick.
The takeaway is not to dismiss the problem. The takeaway is to demand better solutions. Yang’s proposal is a first draft. It needs revision. It needs a mechanism for enforcement. It needs a definition of AI revenue. It needs a global agreement. Until then, it is a talking point. The market will move faster than the policy. AI will continue to displace jobs. The displaced will look for answers. They will not find them in a tax code that cannot be enforced. Logic survives the cold burn. The cold burn is the reality of enforcement. The hype is the proposal. The logic is the impossibility.