Bethany Andres-Beck, a software engineer running for Congress in Massachusetts’ 6th District, used an interview published Wednesday on James Wang’s newsletter Weighty Thoughts to lay out a tax parity plan for automation. She wants government to tax the purchase of AI systems at the same effective rate it taxes wages, closing what she calls a structural discount for replacing workers with software. The proposal is one plank in a wider AI platform that also includes named-human liability for every deployment and a single government-built foundation model meant to blunt monopoly pricing.
Andres-Beck’s number: hiring a worker carries an effective tax rate of roughly 21 percent on the wages paid, she said, while buying a piece of automation to do the same job is taxed at closer to 5 percent because of accumulated depreciation and equipment rules. She is not trying to legally define what counts as a robot. Her framing taxes the output, production, rather than the method that produced it, so a company cannot dodge the levy by describing its software as something other than automation.
That gap is not abstract for anyone running an AI-heavy company. A roughly 16-percentage-point tax advantage currently sits on the side of replacing a role with software, independent of whether the software is actually more productive than the person it replaces. Andres-Beck’s plan would remove that standing subsidy, forcing the build-versus-hire decision to rest on productivity rather than tax arbitrage.
Andres-Beck, who describes herself as a member of the Democratic Socialists of America, opposes government mandates on which models companies may use, a position she attributes to her background in open-source software and technology platforms. Her preferred lever is liability, not model selection: she wants consumer-protection law that names a specific human accountable for each AI deployment, arguing responsibility should attach to how a model is used rather than to the model’s existence.
She also rejects the government equity stakes, so-called golden shares, that Sen. Bernie Sanders has proposed for frontier AI labs. Andres-Beck called the idea “a terrible idea” and compared the likely outcome to Amtrak, a rail operator she said has never inspired much public affection despite decades of federal backing. Her alternative: if pretraining and post-training economics push the industry toward a natural monopoly, government should fund one foundation model that every other lab distills from, rather than trying to referee three or four competing giants. She suggested housing that model at an institution resembling the Library of Congress, reasoning that a system trained on the collective published record of human writing should return value to the public that produced the underlying text.
Her stated worry is pricing power, not capability. Once an AI company becomes the default with no viable alternative, she argued, it can raise prices sharply without a competitive check, a pattern she likens to vendor lock-in in enterprise software. She also framed AI backlash as more geographic than partisan: data center sitings, she said, can draw roughly 2,000 objectors to a single local meeting, a turnout she contrasted with the near-total absence of comparable protest against self-driving cars.
For operators running AI-heavy businesses, Andres-Beck’s platform is worth tracking regardless of how her race turns out. A federal candidate is now campaigning explicitly on erasing the tax gap that currently favors automation over payroll, and on making a named individual, not a model vendor, liable when a deployment goes wrong. Companies that built cost models around today’s tax and liability gap should treat both as assumptions a growing wing of Democratic politics wants to close.
This account is drawn from an interview between James Wang and Bethany Andres-Beck published on Weighty Thoughts on August 19, 2026.
