Tech-Driven Taxes: Privacy-First Filing, Automation-Neutral Incentives, and a Multilateral Path Out of the Digital Tax Standoff

Filing: the privacy failure already happened, and shutting down the alternative didn't fix it

Commercial tax software and paid preparers still handle the overwhelming majority of individual returns. In July 2023, a congressional investigation led by Sens. Elizabeth Warren and Ron Wyden found that H&R Block, TaxAct, and TaxSlayer had installed Meta's tracking pixel on their filing software, sending Facebook a taxpayer's approximate income, filing status, refund amount, dependents, and, in TaxAct's case, calculated federal tax owed, for tens of millions of returns. The investigators called the practice likely illegal. This wasn't a hypothetical risk model; it was documented behavior inside the tax-prep products most Americans already use.

That same industry has a two-decade record of fighting the free alternative rather than competing with it. Intuit sued and spent more than $1.7 million lobbying California legislators between 2001 and 2010 to kill ReadyReturn, a state pilot that pre-filled simple returns from employer and bank data and that more than 90% of participants said saved them time. At the federal level, ProPublica documented that Free File Alliance members, Intuit among them, wrote code to keep their own free-filing pages out of search results. The FTC filed a complaint in 2022 alleging Intuit had deceptively advertised TurboTax as free to consumers who mostly couldn't use it for free, and issued a final order to that effect in January 2024; the Fifth Circuit vacated that order in March 2026, holding the FTC's in-house adjudication of the claim unconstitutional and sending any further enforcement to a federal court instead.

The IRS's Direct File pilot was built without any of that: no ads, no data brokering, nothing to sell. It ran in 12 states for the 2024 filing season (140,803 returns) and grew to roughly 24 states and 30 million eligible taxpayers for 2025, with 90% of GAO-surveyed users rating the experience positively. It also never reached most of them: about 751,000 taxpayers registered, and only about 41% of those filed through it, according to a Treasury Inspector General for Tax Administration (TIGTA) review. In November 2025, Treasury Secretary Scott Bessent announced the program's end, citing cost: roughly $41 million and "at least $138 per return" for tax year 2024. TIGTA's own review of a different fiscal year found the IRS's internally reported cost estimate ($61.2 million) exceeded what the program spent ($16.2 million) by about $45 million, a separate figure on a separate accounting basis that this issue does not attempt to reconcile with Bessent's number, because the two claims don't share a common scope. The Direct File Act of 2026, backed by more than 150 House and Senate Democrats, would restore and permanently codify the program; the administration's own alternative, funded through a $15 million task force created by the same 2025 tax law that ended Direct File, is tasked with getting up to 70% of filers to free filing through the private sector instead.

That fork is the fight that matters. Whichever side wins the argument over whether government or industry should run free filing, neither side's current plan requires that the software touching a citizen's return be free of ad-tech trackers. This issue takes a position underneath that fight: revive a no-tracker government channel for filers who want one, and write the same no-tracker floor into every IRS Authorized e-file Provider, reaching the paid commercial software most filers actually use rather than only the narrower set certified through Free File, so the privacy guarantee doesn't hinge on which side wins.

The tax code's automation bias, and why a robot tax is still the wrong fix

The most-discussed automation-tax proposal in 2025-2026 came from Sen. Bernie Sanders, whose Senate committee staff estimated in an October 2025 report that AI and automation could eliminate close to 100 million U.S. jobs over the next decade, and who has said he plans to formally introduce a tax on companies that replace workers with automation, roughly pegged to the payroll taxes the displaced worker would have generated. OpenAI's own 2026 policy blueprint floated a version of the same idea. Neither proposal has passed, and both face the objection Larry Summers raised the last time a prominent robot tax got serious attention, in 2017: singling out one labor-saving technology for a special tax, when word processors, dishwashers, and electrification all displaced workers too without a dedicated tax, doesn't identify what's different about this case. South Korea's 2017 "robot tax" turned out to be a reduction in existing robotics tax credits rather than a new levy, and the European Parliament rejected an outright robot tax that same year over competitiveness concerns.

Summers and Sanders are arguing about the wrong variable. A 2020 study by MIT economists Daron Acemoglu, Andrea Manera, and Pascual Restrepo, published as an NBER working paper and later in the Brookings Papers on Economic Activity, found that the U.S. tax code's effective rate on labor income runs about 28.5%, while its effective rate on capital invested in equipment and software has fallen to about 5%, largely because of depreciation and expensing provisions enacted between 2002 and 2017. That roughly 23-point gap doesn't ask whether automation is good for productivity; it asks whether the tax code, rather than the market, is why a firm chooses a machine over a hire at the margin. The 2025 One Big Beautiful Bill Act widened that exact gap: it made 100% first-year bonus depreciation permanent for equipment placed in service after January 19, 2025, and raised the Section 179 expensing cap to $2.5 million, on top of the depreciation rules Acemoglu's team had already measured.

This issue's proposal has two parts, and neither is symbolic on its own. The first is disclosure at the point where Congress already legislates the gap: require Treasury and the Joint Committee on Taxation to score the labor-versus-equipment effective-rate gap alongside the revenue estimate every time a bill touches depreciation or expensing, the same way JCT already publishes distributional tables. Disclosure alone would not move a single rate; publishing a gap that Congress is free to ignore changes nothing about what a firm pays at the margin. The second part is the operative one: phase in a ten-year reduction in the employer-side payroll tax rate, funded by narrowing the 2025 law's 100% bonus depreciation and expanded Section 179 cap by an equivalent amount each year, sized to JCT's own published gap estimate. That structure moves rates instead of merely describing them, without requiring Congress to repeal the 2025 law outright in a single vote it will not take. Acemoglu's own modeling found that closing the gap through neutral tax treatment could raise employment by roughly 4% and the labor share of income by nearly a full point; a more modest reform gets a smaller employment gain and, in their model, might still need a supplementary tax on top to fully capture it. This issue treats the phase-in as the mechanism doing that work, and holds a targeted automation tax as a question to revisit only if a decade of narrowing the gap turns out not to be enough.

Digital services taxes: the multilateral off-ramp Washington gave up

The digital services tax fight isn't new. In 2019, USTR's first Section 301 investigation found that France's DST discriminated against U.S. companies, specifically naming Google, Facebook, Apple, and Amazon, and the U.S. ultimately determined a 25% tariff on $1.3 billion of French goods: cheese, sparkling wine, and handbags, industries that had nothing to do with the tech companies the DST targeted. That action was suspended so multilateral OECD talks, eventually split into "Pillar One" (reallocating taxing rights over digital profits) and "Pillar Two" (a global minimum tax), could try to replace DSTs with one negotiated formula instead of a country-by-country fight.

In January 2025, the new administration withdrew the United States from that negotiation on its first day in office. European DSTs, which countries had mostly kept on the books only as leverage to force a multilateral deal, had no remaining reason to stay temporary. In February 2025, the administration directed USTR to revive Section 301 investigations against France, the UK, Italy, Spain, Austria, Turkey, and Canada; by June 2025, Canada rescinded its DST hours before its first payments were due, after the President suspended all trade talks with Ottawa over the tax. The EU's DSTs remain, and the trade deal the U.S. and EU ratified in June 2026, capping most EU tariffs at 15%, explicitly left digital services taxes out of the agreement. The administration renewed a 100% tariff threat against DST-imposing countries the same month. Separately, in January 2026, U.S. negotiators reached a "side-by-side" arrangement largely exempting American multinationals from Pillar Two's minimum-tax top-ups, showing the multilateral channel can still produce results for U.S. firms even after the broader withdrawal.

This issue's position: re-enter the Pillar One-style negotiation the U.S. left, rather than keep running the DST fight one bilateral tariff threat at a time. The Pillar Two side-by-side deal is itself the evidence for this: multilateral negotiation delivered a favorable outcome for American companies within a year of the supposedly abandoned track producing one. Tariff threats protect the same handful of large firms but draft unrelated export industries into the fight, the way French cheesemakers and champagne producers were drafted into a dispute about the digital economy in 2019. A formula that ends DSTs everywhere protects the targeted firms without the collateral industries.