The Uses and Limits of Universalism
Bernie Sanders sees concentrated economic power with unusual clarity. The limit appears at the boundary of who his politics is protecting — and the February 2026 AI pivot shows what happens when that boundary expands to humanity itself.
IThe Instrument at Full Power
Bernie Sanders does one thing better than almost anyone in American politics: he names concentrated economic power without euphemism. Where other politicians say stakeholder, Sanders says billionaire. He asks who owns the system, who profits from it, who works inside it, and who absorbs the cost. That clarity is not incidental to the argument that follows. It is the reason the argument matters.
His record supplies substantial evidence that this framework can protect people who are politically vulnerable. Sanders voted against the Defense of Marriage Act, against the 1996 welfare-reform law, against the USA PATRIOT Act and its reauthorization, and against authorization of the Iraq War. His 2026 AI agenda likewise contains a serious material critique of ownership, worker displacement, surveillance, environmental costs, and the concentration of technological power in a small number of firms and billionaires.
Those facts are not a preface to be politely acknowledged and then ignored. They constrain the thesis. Sanders cannot honestly be described as a politician who simply fails marginalized people, or as someone whose class politics prevent him from recognizing every non-class harm. The pattern is narrower: his politics are most precise when the group he is protecting fits comfortably inside the political subject he already knows how to defend — the worker, the citizen, the Vermont constituent, the public against the oligarch.
IIThe Boundary of Solidarity
Angela Davis identified one version of this problem in 2016. Speaking on Democracy Now, she called the issues Sanders raised “absolutely important” while criticizing what she described as his “economic reductionism”: a framework that, in her view, lacked the vocabulary to treat racism as constitutive of American capitalism rather than as a secondary harm economic justice would eventually solve.
Davis's formulation is useful, but the public record points to something more concrete than an abstract conflict between class and identity. Sanders's solidarity often has a boundary. Inside it, he can be unusually forceful. Outside it, the same analysis can lose resolution. The recurring issue is not whether he cares about people beyond that boundary. It is whose interests define the policy problem when interests collide.
That boundary is not racially neutral in its effects. Current Census Bureau estimates put Vermont at 91.6 percent White alone and 89.4 percent White non-Hispanic, compared with 1.8 percent Black and 2.8 percent Hispanic or Latino. Demography does not establish Sanders's motive. It does matter when policies protect Vermont interests while burdens fall elsewhere on communities that are substantially poorer and less white.
The recurring structure is best described as bounded solidarity: a politics capable of deep solidarity within its recognized constituency, but less reliable when justice requires allowing people outside that constituency to define the problem, impose a cost on the protected group, or remain politically primary rather than being translated into a broader class or universal frame.
IIISierra Blanca: The Boundary Made Visible
In 1998, then-Representative Sanders supported the Texas Low-Level Radioactive Waste Disposal Compact, which allowed Vermont and Maine to send low-level radioactive waste to Texas. The expected disposal site was Sierra Blanca, a poor, heavily Latino border community. Environmental-justice scholar Robert Bullard told the Texas Tribune that the arrangement reflected poorly on Sanders and described forcing the burden onto the community as “business as usual,” not progressive politics.
The same reporting preserves an especially revealing exchange. Sierra Blanca activist Gary Oliver asked Sanders to visit the community. Sanders reportedly declined: “Absolutely not. I'm gonna be running for re-election in the state of Vermont.” Texas regulators eventually rejected the Sierra Blanca site, so the waste was never deposited there. Sanders's campaign also correctly noted that the federal compact did not itself designate Sierra Blanca.
The episode does not establish racial animus, and it does not require one. Vermont had a waste-disposal problem. The anticipated burden was hundreds of miles away in a poor Latino community. When residents asked Sanders to confront that burden directly, his answer returned to the constituency that elected him. “He was representing Vermont” explains the political incentive. It does not erase the racial and economic distribution of benefit and burden.
Sierra Blanca therefore supplies stronger evidence than the softer claim that Sanders sometimes “universalizes” harms. The relevant structure is material: an overwhelmingly white home-state constituency received protection while a distant minority community was positioned to absorb the cost. The record supports that distributional claim. It does not support the stronger claim that Sanders preferred white people because they were white.
IVWho Counts as the Worker?
Immigration exposes the same boundary in labor politics. During the 2007 comprehensive immigration debate, Sanders supported bringing the legislation to the floor but later voted against the final cloture effort on S. 1639. The bill combined legalization and a path toward citizenship with temporary guest-worker provisions Sanders opposed. The Senate record shows the final cloture motion failed 46–53 on June 28, 2007.
Sanders's objection was not simply hostility to immigrants. He repeatedly argued that guest-worker programs could exploit migrant labor while also depressing wages for workers already in the United States. That formulation persisted. In a later Senate statement on immigration reform, Sanders said he wanted comprehensive reform passed while opposing a large expansion of temporary foreign labor because the middle class was working longer hours for lower wages. His more recent H-1B arguments make the ordering explicit: guest workers should receive stronger protections, but corporations should recruit and hire qualified American workers first.
The tension is not that Sanders cares about American workers and therefore cannot care about immigrants. He plainly expresses concern for both. The hierarchy appears when their interests conflict. The protected political subject is the established American worker; the foreign worker is simultaneously recognized as exploited and treated as labor-market access that may need to be constrained.
This is bounded solidarity in its cleanest labor form. Sanders's critique remains anti-corporate throughout. What changes is which worker is allowed to define the remedy.
VWhen Black Constituencies Define the Problem
At Netroots Nation in July 2015, Black Lives Matter activists interrupted the presidential forum and demanded direct attention to police violence and structural racism. Contemporary reporting described Sanders as frustrated and defensive; he invoked his civil-rights record, and his campaign later skipped a planned meeting with some organizers and sent campaign manager Jeff Weaver instead. Sanders then adapted quickly: within weeks he was naming Black victims of police violence explicitly and his campaign was consulting activists on police reform. The Washington Post documented both the confrontation and the subsequent change.
Four years later, the friction returned at the She the People forum in Houston. Before an audience largely composed of women of color, Sanders was asked about white-supremacist violence and about winning the support of Black women specifically. He returned to his civil-rights history, universal programs, and an agenda for “all people.” Audience members objected. The Washington Post described the reaction and the specificity the audience was demanding.
The significant distinction is not between Black people who supported Sanders and Black people who did not. Sanders had prominent Black allies, staff, surrogates, and supporters. Their support is real. The more revealing divide is between Black political actors operating comfortably inside Sanders's existing coalition and Black political actors insisting that their own experience reorganize the agenda. The friction appears when the constituency does not merely ask to be included, but insists on defining the harm.
That is why the familiar defense — Sanders marched for civil rights, Sanders had Black surrogates, Sanders supported universal programs that would materially benefit Black people — does not settle the analysis. Those facts establish solidarity. They do not establish that Black constituencies retained equal power to set the terms when their analysis diverged from his.
VIThe February Pivot
Sanders entered California in February 2026 already concerned about existential AI risk. His February 4 pre-trip statement listed four areas: billionaire control, workers, democracy, and whether AI posed an existential threat to the human species. The chronology therefore does not support a conversion story in which Eliezer Yudkowsky or Nate Soares introduced Sanders to extinction risk.
The chronology does support a pivot in emphasis. Sanders met in Berkeley with Yudkowsky and Soares of the Machine Intelligence Research Institute, alongside other AI figures. Immediately after the California trip, his rhetoric intensified. On February 20 he described the moment as the “most dangerous” in modern U.S. history and warned that policymakers had “not a clue” about the “tsunami” coming from AI. On February 23, his data-center moratorium statement added the possibility that AI could become smarter than humans, escape human control, and produce cataclysmic outcomes.
The frame then became policy. In May, Sanders called for progress toward an international treaty banning superintelligence — an artificial mind smarter than any human and capable of operating independently beyond human control. In September, he and Representative Greg Casar announced the Ban Artificial Superintelligence Act, proposing a permanent prohibition on superintelligent systems and a temporary pause on advanced AI development until a federal regulator established safety rules.
Calling February a pivot does not require claiming causation. Before the trip, extinction risk was one concern among four. After the meetings, loss of human control and superintelligence became a sustained organizing feature of Sanders's AI politics, ultimately producing a dedicated legislative program. The before-and-after change is observable even though the internal mechanism producing it is not.
VIIWhat the Pivot Carries
Sanders's material AI critique remains recognizably his own: ownership, worker displacement, democratic control, electricity prices, land and water use, surveillance, and the distribution of productivity gains. His April 2026 op-ed, “Artificial intelligence is coming for the working class”, is largely organized around those harms. The same period also shows a growing second register: AI as an autonomous intelligence that may outrun human control and threaten humanity itself.
Those two critiques are not interchangeable. The first asks who owns technology, who profits, who loses work, who pays for infrastructure, and who has democratic power. The second accepts a far larger premise: that frontier AI may become an independent civilization-scale actor. That premise is contested, but once accepted it changes the political object. The subject is no longer a worker, a Black community, an immigrant, a woman targeted by a synthetic sexual image, or a neighborhood carrying a data center. The protected constituency becomes humanity.
That enlargement fits Sanders unusually well. The extinction frame identifies billionaire technologists as antagonists while offering him the most universal constituency possible. It does not ask him to surrender his anti-oligarch politics. It allows him to intensify them while adopting the core claim made by the AI-risk ecosystem: the technology is so powerful that ordinary political categories are no longer sufficient.
This is also where the strange convergence with the industry appears. AI builders and AI doomers disagree radically about prescription. Yet both can describe frontier AI as historically extraordinary, transformative, and capable of remaking civilization. OpenAI CEO Sam Altman publicly credited Yudkowsky in 2023 with doing “more to accelerate AGI than anyone else,” including helping make AGI discussable and fundable long before it was mainstream. As documented separately in The Network Behind the Moderate, MIRI's history also includes substantial early support from Peter Thiel.
The convergence does not establish coordination, capture, or shared intent. It establishes a structural possibility: a doomer can sincerely oppose frontier AI while reinforcing the belief that frontier AI is extraordinarily powerful; a builder can reject the doomer's policy while benefiting from the same perception of extraordinary capability; and an anti-oligarch politician can import that premise into governance precisely because the doomer frame casts the oligarch as the villain.
This is the larger significance of bounded solidarity. Sanders's political instrument is sharpest when it knows who must be protected from concentrated power. Sierra Blanca shows what can fall outside that protection geographically. Immigration shows what can fall outside it economically. The Black Lives Matter and She the People episodes show what happens when a constituency insists on defining harm in its own terms. AI presents the inverse move: instead of narrowing the frame toward a specifically burdened population, extinction risk expands it until everyone is inside.
Sanders’s opposition to oligarchy does not prevent his AI framing from enlarging the political importance of the very industry he distrusts. By treating frontier AI as a civilization-scale object of governance, he helps ratify the premise that these machines are important enough to reorganize politics around.
That premise does not belong only to the oligarchs. Taken seriously, it points toward a future they may like much less. A technology built from accumulated human knowledge, labor, language, art, code, and culture does not have to remain the private machinery of a handful of firms. Under different ownership, access, infrastructure, and governance, it could become a tool for breaking concentrations of power rather than deepening them. Sanders himself approaches that conclusion in his proposal for a public AI wealth fund, which argues that the public should hold a substantial ownership stake in the AI economy because its value is built on collective human production.
That possibility gives the mythology around AI a second edge. Human cultures repeatedly produce stories of the liberator: Prometheus bringing fire, Moses leading people out of bondage, the messianic king, the revolutionary machine that finally ends scarcity or domination. AI fits easily inside that old story because it appears to concentrate knowledge, coordination, and productive capacity into a form that can be deployed at scale.
The myth becomes politically consequential when the machine is treated as more than a tool — as something capable of breaking chains human institutions have failed to break. With sufficient public leverage, AI could redistribute forms of expertise and productive power now monopolized by institutions and elites. That is a materially different future from one in which a few firms own the models, the compute, the interfaces, and the gains.
The same premise therefore carries both danger and possibility: these systems may become unusually powerful. Under concentrated ownership, that power can harden hierarchy. Under genuinely distributed control, it could weaken some of the structures hierarchy depends on. The political struggle is ultimately over who holds the lever if that power arrives.
