ANALYSIS: When Sanders and Bannon Sound the Same Alarm About Artificial Intelligence

Introduction: Two opponents discover the same locked door

Washington, September 15, 2026

Two adversaries are moving toward the same warning. Senator Bernie Sanders and former presidential adviser Steve Bannon have both advocated in Washington for much stricter regulation of advanced artificial intelligence.

They did not reconcile. They did not share a common doctrine. They merely acknowledged the same fundamental shift: a few private companies are developing systems capable of transforming work, security, and democracy without a mandate from the people.

Le Monde published their speeches on September 17. This documentary-style coverage reveals more than just an unexpected convergence. It highlights a question that can no longer be ignored: Who actually has the power to slow down these new models when their creators refuse to do so on their own?

The scandal does not lie in their agreement. It lies in the closed door they are pointing to together.

The Absent Power Behind the Speeches

The political stage attracts attention. Yet the real action lies elsewhere. Two opposing traditions acknowledge that democratic institutions lag behind technology, behind investments, and sometimes behind the damage.

Sanders speaks of popular sovereignty. Bannon invokes national sovereignty. Their paths are already diverging. But both acknowledge that industrial competition imposes its timeline on citizens, rather than the other way around.

This convergence does not prove that their solutions are compatible. It proves that the unease now extends beyond partisan lines. And yet, Washington continues, above all, to celebrate speed.

When the extremes of the debate hear the same siren call, the center can no longer remain silent.

Technology advances while permission is pending

Models launched before public debate

New systems are designed, trained, and then deployed based on business decisions. The public weighs in afterward, often once the systems are already in place. Democratic consultation then becomes a belated remedy, not prior authorization.

Sanders emphasized this disempowerment in his September 15 speech. According to him, the American public has had virtually no say in the choices that are already shaping their jobs, their privacy, their education, and their information environment.

The problem therefore goes beyond technical safety. It concerns legitimacy. A machine can function in accordance with its designers’ expectations while transforming society according to rules that no one has actually voted on.

Unauthorized innovation can become a government without a ballot.

Consent Postponed Until After September 2026

The timeline speaks for itself. Companies are investing. Infrastructure is taking shape. Models are gaining capacity. Citizens then discover the local cost: constant noise, electricity demands, pressure on real estate, or job insecurity.

Who gave their consent?

The question remains open because the market responds through consumption, not through sovereignty. Yet refusing a product does not always allow one to escape the automated decisions made by an employer, a government agency, or a platform.

The “accept” button will never replace a collective choice about how society is organized.

Their consent begins exactly where it breaks down

Two assessments made on the same day

Sanders and Bannon both denounce the concentration of technological power. They target leaders with immense capital, direct political access, and an exceptional ability to define the vocabulary of the debate.

The progressive senator wants to place this power under democratic control. Donald Trump’s former advisor favors an executive, national response directed against China. Their shared assessment, therefore, does not produce any common policy.

This difference matters. It prevents their successive appearances from turning into an alliance. They momentarily occupy the same crossroads, then go their separate ways.

A shared sense of alarm does not build a shared home.

The divide revealed in Washington

Sanders envisions federal regulation, a conditional pause, and international agreements. Bannon considers Congress too slow and calls for swift presidential action accompanied by a radical hardening of stance toward Beijing.

The former seeks a global framework. The latter wants to win a domestic showdown. Their disagreement shows why the fear of losing control is not enough to establish legitimate control.

And yet, their occasional alignment serves a political purpose. It deprives the accelerationists of a convenient refuge: the idea that all caution belongs exclusively to one ideological camp.

The debate has just shed its label. It must now address the substance.

Speed serves as an argument for those who choose it

The U.S. competition in September 2026

The White House presents U.S. technological dominance as a strategic priority in the face of China. Following this logic, slowing down a national laboratory risks ceding ground to a competitor with fewer constraints.

This argument carries real weight. No Western government can ignore the economic, military, and scientific applications of AI. But it easily turns any rule of caution into a threat to national power.

The trap closes in this way: each new capability justifies the next. Each investment makes pausing more costly. Each rival serves as an excuse for the next deployment.

The race then becomes its own justification.

The Timeline Imposed by Research Labs

Public authorities debate in months. Companies think in terms of product cycles and market windows. This asymmetry often forces lawmakers to regulate technology that is already outdated.

Faster.

This word may seem technical. It is political. Acceleration determines which risks will be tolerated, which industries will undergo the transition, and how much time institutions will be given to understand what they must regulate.

Speed is not neutral. It distributes power before the law distributes responsibilities.

The actual product being sold has a simple name

The industry’s promise since 2022

Companies market tools. They also sell a certainty: adoption is inevitable. This rhetoric reduces policy to merely facilitating a future presented as already decided.

Yet no industrial trajectory is natural. Infrastructure requires permits. Companies benefit from tax rules. Chips circulate under licenses. Technological inevitability rests on very concrete public decisions.

What no one says often enough is that the supposedly autonomous future depends on electrical grids, capital, workers, data, and legal protections organized by the community.

The “inevitable” is often a policy that refuses to say its name.

The Word “Control” After September 15

Sanders uses “control” to refer to democratic sovereignty over corporations. Bannon uses it to defend national authority and counter Beijing. Industry leaders, on the other hand, speak of technical control over models.

These meanings are not interchangeable. A system can remain technically manageable while reinforcing massive economic concentration. It can also be regulated at the national level without granting more power to workers or citizens.

Who controls the control?

That is the question buried beneath the rhetoric. Surveillance without sovereignty gives nothing back to the public.

Concerns have already spread beyond the laboratories

The Pew survey published on June 17, 2026

A Pew Research Center survey, conducted in February 2026 and published in June, indicated that 40 percent of American adults anticipated a negative impact of AI on society over the next twenty years.

Only 16 percent expected a positive impact. Furthermore, 63 percent felt that the technology was advancing too quickly. These findings predate the Washington conference, but they shed light on its resonance.

Concern, therefore, is no longer limited to a handful of security experts. It extends to a population that perceives the transformation but lacks a common language to manage it.

The public does not necessarily reject the technology itself. It refuses to be presented with a fait accompli.

Eroded Trust Ahead of the Elections

A concerned public does not automatically demand a ban. It often calls for guarantees, transparency, and avenues for recourse. However, a lack of answers can turn uncertainty into lasting opposition.

And yet, part of the official discourse continues to equate caution with a fear of progress. This reflex disregards what citizens are truly expressing: the fear of losing all control over their own future.

The rumblings remain diffuse. They can be heard in discussions about jobs, schools, data centers, and privacy. Then they find their way to the ballot box.

A democracy can ignore a single concern. It cannot ignore its accumulation indefinitely.

Fear of the machine sometimes masks fear of the owner

Concentrated economic power in 2026

Sanders places AI within a long-standing narrative: that of wealth created collectively and then captured by a few owners. Models learn from texts, images, research, and code produced by generations of humans.

His analysis, therefore, does not focus solely on an uncontrollable system. It targets the ownership of infrastructure, data, and profits. The machine is a source of concern, but its owner already calls the shots.

Bannon, too, attacks the tech giants’ vast fortunes. His framework, however, remains nationalist. He does not seek the same kind of redistribution. The word “oligarchy” brings their discourses closer together, without giving them the same substance.

The machine does not accumulate wealth on its own. Someone is always cashing in behind the screen.

The Proposal Put Forward in June 2026

On June 18, Sanders presented a proposal for a sovereign wealth fund dedicated to AI. The proposal notably called for public ownership stakes in the largest U.S. companies in the sector.

This earlier initiative helps explain his September speech. He is not merely calling for a safety barrier. He wants to shift ownership, decision-making, and the distribution of value.

This framework distinguishes his proposal from Bannon’s. One asks who owns it; the other asks which country dominates. These questions may overlap, but they do not lead to the same world.

Confusing popular sovereignty with national power would obscure the core of the disagreement.

A pause is only valid if the one who can order it does so

The Sanders-Casar Proposal of September 3

Sanders and Democratic Representative Greg Casar announced on September 3 a proposal to ban artificial superintelligence and temporarily suspend certain advanced developments.

The bill calls for the creation of a federal authority tasked with establishing rules, reviewing models, and monitoring dangerous capabilities. It also envisions strong penalties for violations.

It remains a policy proposal, not a law in effect. Its scope, however, signals a major shift: the call for a slowdown is becoming a concrete legislative issue, no longer just a petition from researchers.

Simply saying “pause” is not enough. We need an institution capable of declaring such a pause and enforcing it.

The Problem of Definition in September

What should be suspended? A model that exceeds a computational threshold? A system with certain capabilities? A specific training process? A category as fluid as advanced AI defies simple boundaries.

A definition that is too broad may block useful applications. A narrow definition can be circumvented. A standard based on capabilities requires independent evaluations and technical access that companies are reluctant to share.

Then comes the dizziness…

A poorly defined pause becomes either a blanket ban or a door left ajar.

Companies sometimes call for the very rule they fear

Industry Calls from September 2026

Several industry leaders and companies have acknowledged the need for mechanisms to slow down certain developments. This openness marks a break with the doctrine that self-regulation is always sufficient.

Nevertheless, it must be examined without naivety. A large company may support regulations because it fears a genuine risk. It may also favor costly obligations that smaller competitors will be unable to meet.

Security and consolidation are not mutually exclusive. A single measure can protect the public while strengthening already dominant players. Regulation must therefore monitor both risk and the market.

When a giant calls for a barrier, we must verify on which side it owns the land.

The Risk of a Regulatory Cartel

Bannon emphasizes the possibility that major tech companies might design a regulatory framework that favors their own interests. His suspicion touches on a classic problem: regulated entities often help draft the very standards that govern them.

Their expertise is indispensable. Allowing them to control the process would be dangerous. Regulators need to understand complex systems without becoming dependent on the companies they are tasked with overseeing.

In other words, regulations designed exclusively with industry leaders can legitimize their competitive advantage. The safeguard then becomes a wall protecting the established players.

Regulations must govern these companies. They must never belong to them.

China Becomes Both the Excuse and the Real Problem

The U.S. Doctrine of September 2026

Competition with China shapes U.S. policy. Washington wants to preserve its lead in chips, infrastructure, and models. Beijing is pursuing its own economic, military, and administrative capabilities.

This rivalry cannot be dismissed with a single gesture. A moratorium limited to democracies might merely shift certain research efforts rather than halt them. National security, therefore, is fully part of the equation.

But the rivalry also gives rise to a narrative that neutralizes any form of oversight: if we slow down, the other side will win. This statement shuts down the debate before the risks are even examined.

China is a strategic threat. It must not become an unlimited license granted to every Western laboratory.

Two Incompatible Responses on September 15

Sanders raises the possibility of international agreements comparable, in function, to arms control mechanisms. Bannon, on the other hand, proposes a much more drastic economic and technological isolation of China.

The first approach requires a minimum level of trust, inspections, and reciprocity. The second envisions a rapid separation of deeply intertwined economic systems. Neither guarantees straightforward implementation.

And yet, an international approach is essential. A powerful model can be developed in one country, funded elsewhere, and then deployed everywhere. A national border is not enough to contain digital capabilities.

The risk is global. Decision-making remains confined to competing capitals.

The word “human” unites because it decides nothing

The Future of Life Institute’s Statement

The Pro-Human Assembly was organized by the Future of Life Institute. Its program advocates for human control, corporate accountability, the protection of freedoms, and the limitation of concentrations of power.

These principles explain the diversity of the speakers. They provide a common ground before disagreements arise. Everyone can affirm that AI must serve people rather than replace or govern them.

But the term “human” remains broad enough to encompass contradictory projects. It specifies neither the ownership of models, nor safety thresholds, nor the role of labor, nor the powers of the regulator.

A principle unites people as long as it does not yet allocate losses and obligations.

The Battle Over Definitions in 2026

For an employee, a “pro-human” AI might mean avoiding automated layoffs. For a company, it might refer to a productive assistant. For a government, it could become an instrument of power.

For a parent, the issue sometimes begins with a screen, a synthetic voice, or a school that adopts a tool without a clear explanation. The hum of the server remains far away. The decision, however, enters the home.

Who defines dignity when interests clash?

Humans must not become mere moral window dressing for an unchanging industrial decision.

Work will be the first arbiter of the promise

Jobs at the Heart of Sanders’ Message

Sanders lists the transformation of work among his top concerns. Generative systems are already automating certain intellectual tasks, accelerating production, and changing the expectations placed on employees.

The consequences remain uneven. Some workers save time. Others see their jobs fragmented, monitored, or devalued. Companies can distribute productivity gains or use them to cut jobs.

Technology does not determine how these gains are shared. Contracts, unions, taxation, competition, and labor laws will—or their absence will.

The question is not merely what the machine can accomplish. It is what the owner will do with that capability.

The Transformed Workplace After 2026

Automation doesn’t always feel like a door closing. It can start with a task being eliminated, a target raised, a position left unfilled, or an evaluation generated by an opaque system.

The keyboard is still there. So is the chair. Yet autonomy is shrinking. The pace is picking up. The employee must control the tool while fearing that this tool will then measure their own usefulness.

This is the tangible feel of change: a colder, continuous, almost silent pressure. It doesn’t wait for the hypothetical arrival of a superintelligence.

The future of work may fade away through small tasks before it disappears in the headlines.

Democracy can be automated without being abolished

Political funding denounced in September

Sanders called on AI executives to stop funding electoral organizations designed to sideline candidates who support stricter safeguards.

His political accusation highlights a contradiction: calling for clear regulation while using considerable financial power to influence those who will draft it. Here, the technical debate intersects with the practical economics of U.S. elections.

A company does not need to govern directly. It can fund, advise, negotiate, and define the terms considered reasonable. Power becomes diffuse—and thus harder to challenge.

Democracy does not always die out. It may simply begin to speak the language of its financial backers.

Public Decision-Making Under Electoral Pressure

As the U.S. midterm elections approach, concerns about data centers, jobs, and energy consumption are taking on increasing electoral significance.

Politicians are now discovering an issue that has long been treated as an industry-specific matter. Citizens are no longer asking only what levels of performance are possible. They are asking why their community should bear the costs.

But if the answer comes only because the election is approaching…

Better late protection than none at all. It nevertheless reveals just how long the public was kept in the dark.

The immediate danger does not have to become absolute

Risks Discussed in September 2026

Discussions mention systems that could circumvent technical limitations, facilitate cyber intrusions, or contribute to dangerous research. These scenarios require independent verification and proportionate responses.

The debate becomes fragile when any uncertainty is treated as certainty. It also becomes fragile when the lack of definitive proof is used to justify inaction. Between panic and denial lies the demanding work of precaution.

I acknowledge a personal fear here: that the gravity of the issue will be undermined by the exaggerations of its own advocates, and then exploited by those who want no rules at all.

An exaggerated alarm is easy to silence. A precise alarm forces us to open the door.

The damage already visible today

We don’t need to wait for intelligence beyond human capacity to take action. Automated errors, surveillance, deceptive imitations, and economic concentration are already part of the present.

The contrast is striking. Decision-makers discuss an extreme future. Meanwhile, ordinary decisions are entrusted to systems whose workings the people affected rarely understand.

Yesterday, the promise was assistance. Today, the issue has become dependence.

This distant scenario must never overshadow the person who is already trapped facing a decision with no recourse.

Congress is chasing a moving target

A proposal that still lacks a majority

The Sanders-Casar bill must navigate a divided legislative system. Industrial interests, partisan divisions, and national security concerns complicate any blanket suspension.

This institutional weakness partly explains Bannon’s call for executive intervention. It does not make this path preferable. A swift presidential decision can be reversed just as quickly by a subsequent administration.

Congress’s slowness sometimes protects against arbitrariness. It becomes dangerous when technology changes faster than our capacity for expertise and oversight.

The democratic process is slow. It must become competent without becoming authoritarian.

The Institutional Choice Beyond Washington

Sustainable regulation requires inspectors, public researchers, computing resources, protections for whistleblowers, and real access to the systems being evaluated.

Passing a law without funding its implementation would create a facade. The regulatory body would bear an imposing name, while the labs would retain the engineers, the data, and the computing power.

Administrative silence.

A law without technical capacity leaves the key on the table for those it purports to oversee.

Convergence creates a window but not a way out

A coalition of opposition formed on September 15

Progressives, conservatives, religious leaders, union representatives, and security experts can agree on what they reject: unchecked concentration of power, opaque systems, and a society transformed without consent.

This coalition remains negative. It knows how to say no. It does not yet know how to define the distribution of benefits, global governance, or the acceptable limits of automation.

Yet its value remains real. It expands the political space for caution. It makes possible a conversation that each side might have dismissed when it came solely from the other.

The window has just opened. There is no guarantee that anyone will know how to build a way out.

The Risk of Co-optation Starting in 2026

Each side can use this concern to bolster its own agenda. Progressives may reduce the entire issue to property rights. Nationalists may reduce it to China. Corporations may reduce it to technical security.

All these dimensions matter. None is sufficient on its own. A serious response must simultaneously address security, competition, labor, civil liberties, sovereignty, and the distribution of wealth.

Otherwise, the term “regulation” will become a smokescreen. Everyone will invoke it. No one will relinquish power.

Regulation begins only when someone finally agrees to relinquish a prerogative.

Conclusion: The machine is already raising a question of sovereignty

The Verdict of September 17, 2026

The publication of Sanders’ and Bannon’s speeches reveals a concern that has become bipartisan. Above all, it shows that the central conflict now pits democratic decision-making against the private agenda of the tech labs.

The two men do not offer a unified perspective. Sanders seeks popular and international oversight. Bannon favors executive authority and confrontation with China. Their disagreement precludes any easy celebration.

But their shared diagnosis shatters a long-standing defense. Demanding limits no longer means rejecting technology. It means refusing to let its owners become the sole architects of our collective future.

Caution is not a step backward. A step backward is voluntarily giving up the right to decide.

The door that remains closed tonight

We will need to identify the relevant models, develop independent assessments, fund the regulatory body, and negotiate with allies. We will also need to stand up to China without copying its concentration of power.

The task seems immense. It is. Yet the alternative would be far more brutal: to let investments already made serve as a mandate, and then call what was never chosen “adaptation.”

Two adversaries have heard the same alarm. Behind the door, the machines are advancing. In front of it, democracy is still searching for the key.

The final human check begins with a human decision: to reclaim the key.

When Sanders and Bannon Sound the Same Alarm on Artificial Intelligence

Who should have the power to halt a model before its deployment makes that decision nearly impossible?

By Maxime Marquette, Columnist

Columnist’s Transparency Statement

Editorial Stance

I am not a journalist, but a columnist, analyst, and expert. My expertise lies in observing and analyzing the geopolitical, economic, and strategic dynamics that shape our world. My work consists of dissecting political strategies, understanding global economic trends, contextualizing the decisions of international actors, and offering analytical perspectives on the transformations that are redefining our societies.

I do not claim to possess the dispassionate objectivity of traditional journalism. I claim analytical clarity, rigorous interpretation, and a critical reading of events—without on-the-ground presence or fabricated personal accounts.

My perspective is openly Western and supportive of Ukraine’s defense. I acknowledge this rather than hiding it behind a facade of neutrality, so that readers can take it into account and judge for themselves.

Methodology and Sources

This text maintains a distinction between verified facts and interpretive analysis. Factual information comes exclusively from verifiable sources and is cross-checked before being included.

Primary sources: official press releases, public statements, institutional reports, reputable news dispatches, and verifiable data.

Secondary sources: reputable media outlets, specialized publications, analyses from established institutions, and sector-specific reports. No figures are included based on a single source.

Statistical, economic, and geopolitical data come from official institutions when available and are cross-checked. When two reliable sources differ, the discrepancy is noted rather than smoothed over.

Nature of the Analysis

Analyses, interpretations, and perspectives constitute a critical and contextual synthesis based on available information, observed trends, and verifiable expert commentary.

My role is to interpret the facts, contextualize them, and give them coherent meaning within the framework of contemporary geopolitical, economic, and strategic dynamics.

Any subsequent developments may alter the outlook presented.

ANALYSIS: When Sanders and Bannon Sound the Same Alarm About Artificial Intelligence