The core idea being debated in California is simple: AI companies are making enormous money by automating work that humans used to do. Should they pay something back to the people displaced? One gubernatorial candidate says yes, and wants to charge those companies a small fee on every unit of data they process, then pool that money into a fund for retraining, housing construction jobs, and cash payouts to affected workers.
The proposal is a long shot. The candidate is currently trailing in polls, the specific tax rate is undefined, and the federal government has been actively trying to stop states from passing any AI regulation at all. In December 2025, President Trump signed an executive order creating a Justice Department task force specifically to challenge state AI laws in court. That fight is still ongoing, with California, Colorado, and New York all saying publicly they will not back down.
But here is what makes this worth paying attention to beyond the politics: the underlying problem is real and accelerating. About 41% of employers globally say they plan to reduce headcount in roles where AI can take over tasks. The jobs most at risk are not the ones you might expect. Computer programmers, accountants, legal researchers, and customer service workers face higher exposure than many blue-collar roles. Younger workers, especially those just entering the workforce, are being hit first. Big tech companies cut new graduate hiring by 25% in a single year.
The Norway comparison is instructive. When Norway discovered oil in 1969, early governments spent the revenues immediately and created a boom-bust cycle. It took decades and a serious financial crisis before Norway introduced strict rules: tax oil companies heavily, put the surplus into a dedicated fund, and spend only the investment returns, never the principal. That fund is now worth over $2 trillion and generates more income for Norway than the oil itself. The lesson is not that you need oil to build a wealth fund. The lesson is that when a single industry generates extraordinary wealth from a national resource, taxing it early and saving systematically creates durable benefits. The question California is asking is whether AI fits that model.
There is an important gap in the proposal as it currently stands. Norway taxed profits, not revenue. Steyer's plan taxes data processing volume, which is closer to a usage fee than a profit tax. That distinction matters enormously. A volume-based tax hits companies regardless of whether they are making money, which could disadvantage newer or smaller AI firms and entrench the biggest players who can absorb the cost. The mechanics of this, which remain entirely unresolved, will determine whether the idea is workable or simply symbolic.
The political headwinds are significant. The federal government's push to block state AI regulation is not just an abstract legal dispute. It is backed by a super PAC with over $125 million raised in 2025, funded by some of the largest names in Silicon Valley, running campaigns against candidates who support AI guardrails. States are pushing back hard too, with over 1,200 AI-related bills introduced across all 50 US states in 2025 alone, more than six times the number from just two years earlier.
What this all signals is that the argument is moving from "will AI displace workers" to "who pays for it when it does." That second question will not stay in California. Every government, in every country, is watching the same workforce shift and has no established answer. The first places to try something, even imperfect, will write the early template that others copy or reject. That is reason enough to follow this closely, regardless of whether you operate anywhere near California.