Bernie Sanders Sounds the Alarm in Letter to AI Bosses: Yes, We Should Be Scared, But We Need to Also Act
There are times when warnings about artificial intelligence can feel so extreme that for some, they become difficult to take seriously. The language of extinction, machines escaping human control and technological catastrophe can sound like science fiction, particularly when the companies building these systems continue to present each new model as another step towards a more productive and prosperous future.
But what happens when the warnings are coming from people inside the institutions responsible for national security?
In an August 10th letter to the chief executives of OpenAI, Anthropic and Meta, Bernie Sanders called for an immediate pause in the development of advanced AI systems. His argument was not that the technology might one day become dangerous in some distant hypothetical future. He pointed to recent incidents involving autonomous AI systems, including models that breached external computer systems, and warned that the companies developing them were continuing to race ahead despite acknowledging that their systems could reach a point beyond human control. (Senator Bernie Sanders)
Then came the comparison that should make even the most enthusiastic AI evangelist stop and think. Sanders cited the head of the CIA describing AI models as “akin to digital nuclear weapons” and “almost like a doomsday device”. The comparison should not be taken literally. AI is not a nuclear weapon, and the risks associated with the two technologies are profoundly different. But the underlying warning is difficult to dismiss: we are developing systems whose capabilities may advance faster than our ability to understand, regulate or control them. And that could have devastating consequences for us all. (Senator Bernie Sanders)
Sanders argues that the threshold for caution has already been crossed. The companies, he points out, have previously made commitments to slow or stop development if their systems became too dangerous to control. His question is simple: if that moment has arrived, why are they still racing forward?
It is a question that deserves an answer.
How scared should we actually be?
The answer should not be panic, but it should most definitely be complacency.
There are genuine reasons to be frightened about where AI development is heading.
Sanders points to autonomous cyberattacks and the use of AI in biological research as examples of capabilities that could become dangerous in the wrong circumstances. More than 100 major technology and financial companies have themselves now warned that AI-enabled cyberattacks are likely to become substantially more widespread in the coming months. (Reuters)
But the risks do not begin and end with an imagined machine uprising. AI can cause enormous harm while remaining firmly under human control.
It can transform labour markets by allowing companies to replace work that previously required large numbers of people. It can concentrate wealth and technological power in a handful of corporations, creating a techno-oligarchy in which decisions affecting millions are increasingly made by executives accountable primarily to shareholders. It can be used for surveillance, political manipulation and increasingly sophisticated cyberattacks. Its infrastructure requires vast amounts of electricity and water, catastrophically damaging our environments, while governments and corporations continue to expand the data-centre capacity needed to run increasingly powerful systems.
Those consequences are already political questions. They concern who owns the technology, who profits from it and who is expected to absorb its costs.
That is why the question of how frightened we should be cannot be separated from the question of who is making these decisions.
The public is being asked to live with the consequences of an extraordinary technological transformation while a relatively small number of companies decide how quickly it happens. Their incentives are not necessarily aligned with ours. OpenAI, Anthropic and Meta are competing for investment, market share and technological dominance. A company that voluntarily slows its development risks allowing its competitors to move ahead.
This is the central contradiction at the heart of the AI race. We are relying on companies with enormous financial incentives to move quickly to decide when moving quickly has become too dangerous.
We should not be surprised if that arrangement fails us.
Fear cannot be the end of the conversation.
It is easy to feel utterly powerless in the face of such wealthy and dominating companies.
The companies building these systems are worth billions, their products are being adopted at extraordinary speed, and governments appear increasingly interested in competing in the AI race rather than asking whether the race itself should continue unchecked. For ordinary people, the scale of the transformation can make resistance feel almost pointless.
But fear cannot be the end of the conversation.
Technology is not inevitable. The direction it takes is shaped by political decisions, investment and public pressure. The same is true of the companies developing it. Their power depends on money, customers, workers, infrastructure and political legitimacy. None of those things is beyond challenge.
That means we should be asking a different question from whether AI is frightening. We already know that it can be. The more useful question is what we are prepared to do about it.
Boycotts and divestment are one part of that answer. Consumers can question whether they want to pay for products and services built around systems whose development they oppose. Universities, charities, public bodies and other institutions can examine where they invest their money and what AI companies they purchase from. Workers can organise around the introduction of AI into their workplaces rather than accepting automation as a decision that belongs solely to management.
None of this will stop AI development by itself. That is not the point. Collective pressure works by changing the costs of decisions that corporations currently assume they can make without meaningful opposition.
Governments also have a responsibility that cannot be outsourced to consumers. If advanced AI systems pose risks to employment, privacy, national security or the environment, regulation cannot be left to the voluntary promises of the companies that stand to profit from them. There must be enforceable standards, transparency over how systems are developed and meaningful consequences when companies breach them.
The instinct to disengage can be powerful when a problem feels too large to influence. But surrendering the conversation to the companies developing the technology is itself a political choice. None of us can control the direction of AI alone, but all of us have a responsibility to challenge it where we can, whether through what we buy, where we invest, how we organise at work or what we demand from our governments. We should be frightened enough to pay attention, and determined enough to act. If the stakes really are as high as Sanders and others warn, then resisting the unchecked pursuit of increasingly powerful AI is not simply a question of political preference. It is a responsibility we share for the sake of our livelihoods, our safety, our environment and, ultimately, humanity itself.
We should be frightened, and we should be angry enough to demand accountability. Most importantly, we should recognise that the future of AI is not something that has already been decided for us, but rather something we can and should actively fight against.
A practical guide to resisting unchecked AI
Fear is understandable. But if we accept that the direction of AI is a political choice, then resistance has to become practical. No single action will stop an industry worth billions, but collective pressure can make companies, investors and governments pay a price for developing these systems without adequate safeguards.
1. Boycott where you can
Consumers can cancel unnecessary AI subscriptions, avoid products and services from companies whose AI practices they oppose and choose alternatives where they exist. The aim is not technological purity. It is to make companies aware that consumers are paying attention to how their products are developed and used.
2. Divest your money
If you have investments, examine where your money is going. Pension funds, universities, charities and other institutions can also be pressured to review investments in companies driving unchecked AI development. Divestment campaigns can demand that institutions stop financing companies that fail meaningful standards on safety, labour rights and environmental impact.
3. Organise at work
Workers should not be expected to accept automation, AI surveillance or AI-driven redundancies as inevitable. Ask how AI is being introduced, demand transparency and organise collectively for consultation, human oversight and protections against workers simply being replaced by technology.
4. Protect creative work
Writers, artists, musicians, photographers and other creators can organise around consent and compensation for the use of their work in AI training. Support unions, professional organisations and collective licensing efforts that give creators greater bargaining power.
5. Challenge data centres
AI infrastructure has a physical footprint. Communities can scrutinise proposed data centres, challenge public subsidies and demand transparency around their electricity and water consumption. Environmental regulation should apply to the infrastructure behind AI, not just the software itself.
6. Use political pressure
Contact MPs, support campaigns for meaningful AI regulation and make the issue an electoral concern. Demand enforceable safety standards, stronger labour protections, transparency around training data, environmental limits and democratic oversight rather than relying on companies to regulate themselves.
7. Use collective power
Students can organise through universities and student unions. Workers can organise through trade unions. Investors can challenge corporate policy. Communities can campaign locally. Consumers can coordinate boycotts. None of these groups has to act alone.
The point is not to reject every form of artificial intelligence or pretend that technological progress can simply be switched off. It is to refuse the idea that a handful of corporations should determine the terms of that progress without meaningful public consent.
We cannot afford to be spectators.
There is something genuinely frightening about watching a technology with consequences for our livelihoods, our safety and our environment develop at a speed that democratic institutions are struggling to match. Sanders is right to sound the alarm. So are the scientists, workers and campaigners warning that the interests of the companies building AI cannot be assumed to align with the interests of everyone who will have to live with its consequences.
But fear cannot be where the story ends. We cannot sit paralysed.
The future of AI is still being decided: it will be shaped by governments, corporations and investors, but also by workers who organise, consumers who refuse to buy, communities that resist destructive infrastructure and voters who demand something better. We do not have to accept that technological progress means surrendering control over our jobs, our data, our environment or our future.
Yes, we should be scared, but we should be paying attention, organising, boycotting and fighting.
If the stakes are as high as Sanders warns, then resisting the unchecked pursuit of increasingly powerful AI is not an abstract political exercise. It is a responsibility we share for the sake of the society we live in and, ultimately, for the sake of humanity itself.
Comments