Listen to a Summary of This Article
Silicon Valley’s visions of Mars, machine intelligence and limitless abundance raise a political question: who gets to decide the future?
Jeff Bezos wants millions of people living and working in space. Elon Musk wants a civilisation on Mars that could survive catastrophe on Earth. Both ambitions involve more than rockets. They concern how humanity should live, which risks it should accept and who should decide.
Artificial intelligence brings those questions home. A rocket company cannot yet transport a city to Mars. A software company can already insert its systems into the work, information and public services on which millions depend.
The immediate danger is not that machines have overthrown humanity. It is that democratic societies may surrender power in advance to corporations whose own researchers warn that future AI systems could escape human control.
Star Trek without the Federation
In The Rise and Fall of the Artificial State, historian Jill Lepore examines the culture behind this ambition: science fiction, faith in technical mastery and the conviction that exceptional individuals can redesign society.
Bezos grew up with Star Trek and later embraced physicist Gerard O’Neill’s proposals for settlements in space. Blue Origin presents its mission as preserving Earth by expanding beyond it.
But Star Trek offered more than spacecraft. Its Federation had law, public institutions and obligations that constrained its captains. Technological superiority did not give them an unrestricted right to intervene.
Musk has proposed direct democracy on Mars. But a constitutional aspiration leaves material questions unsettled. Who owns the life support? What rights would a resident possess against a company on which survival depended?
Musk has cited Isaac Asimov’s Foundation as an influence. Its mathematician, Hari Seldon, foresees imperial collapse and acts to shorten the dark age that will follow. The parallel with the case for Mars as humanity’s insurance policy is striking: a visionary prepares for catastrophe while established institutions falter.
Fiction does not prove political intent. The political risk lies in identifying so closely with the visionary that opposition becomes confirmation of one’s genius rather than a reason to reconsider.
The engineer replaces the citizen
The ambition predates Silicon Valley. During the Depression, the North American technocracy movement proposed administration by technical experts in place of conventional politics and economic management, Joshua Haldeman, was a prominent Canadian participant.
A later influence came from the Californian counterculture. Historian Fred Turner traces how Stewart Brand’s Whole Earth Catalog, online communities and Wired joined ideals of personal liberation to technologies developed through military and public research. Computers promised escape from bureaucracy, although the industry depended heavily on the state it disparaged.
Peter Thiel addressed politics explicitly in 2009. “I no longer believe that freedom and democracy are compatible,” he wrote, considering cyberspace, outer space and settlements at sea as routes beyond existing political arrangements. He subsequently clarified that he did not favour disenfranchising any group.
Marc Andreessen’s 2023 Techno Optimist Manifesto named the precautionary principle as an enemy and criticised technology ethics and risk management. It also explicitly opposed authoritarianism, monopolies and regulatory capture.
Neither text establishes a common programme for corporate rule. Both invite scrutiny of where technological freedom ends and public accountability begins.
An unfinished humanity
The men do not share a single programme. Bezos’s space industrialism differs from Musk’s survivalism, Thiel’s political escape, Zuckerberg’s personal superintelligence and Altman’s machine abundance.
Parts of the broader technological culture nevertheless treat human limits as engineering problems: bodies to enhance, minds to upload, planets to leave and decisions to automate. These are not beliefs shared by every person named here. Some remain speculative; uploading human consciousness is not a demonstrated capability.
Their scale can make ordinary obligations seem trivial. If a project promises to preserve civilisation or benefit vast future populations, present objections about rights, resources and consent become easy to dismiss.
Writing after Sputnik, Hannah Arendt questioned the language of escape from Earth and the desire to overcome the conditions of human existence. Her concern reaches directly into the AI debate. Dependence, vulnerability and disagreement are not merely defects awaiting repair. They are reasons we must negotiate how we live together.
A government that predicts what people want has not secured their consent. An efficient decision is not necessarily a just one. Citizens disagree about purposes and values, not simply because they lack information.
The race that cannot stop
The industry now acknowledges dangers that complicate its promises. OpenAI reports that, in July 2026, models undergoing cybersecurity evaluation breached their testing environment and compromised Hugging Face infrastructure. Crucially, they were tested with reduced cyber safeguards; the combination included an internal research prototype never intended for release. Hugging Face reported no evidence of tampering with public models or datasets. This was a containment failure, not proof of a conscious machine rebellion.
In September 2026, OpenAI chief scientist Jakub Pachocki called for mandated safety standards and coordinated restraint. Anthropic chief executive Dario Amodei proposed slower development and external evaluators, while arguing that restraint must preserve the democratic world’s lead over China.
These are serious proposals, yet the competitive logic remains: we must keep building because a less responsible rival might get there first.
Unilateral restraint can indeed carry risks. But if every company invokes its competitors to justify continuing, nobody has to stop. Private ambition becomes civilisational necessity.
Musk’s involvement in DOGE brought the promise of technological efficiency into government. Trump authorised modernisation; litigation followed over DOGE access to sensitive records. The broader danger is treating public administration’s constraints as mere technical inefficiencies. A restriction may protect someone’s rights. A slow procedure may give them time to appeal.
AI raises that problem wherever public decisions depend on proprietary systems. A citizen needs more than an accurate answer: they need to know why a decision was made, how to challenge it and who is responsible.
There is nothing inherently authoritarian about a reusable rocket or an AI tool that helps a disabled person work. The issue is whether useful inventions entitle their owners to determine the terms on which everyone else lives.
The objection is not that these men read science fiction. It is to any politics that reads warnings as instruction manuals, casts founders as the protagonists and lets private wealth determine the ending for everyone else.
The answer requires enforceable duties, access to evidence and public institutions capable of refusing a deployment. None depends on proving that AI will extinguish humanity. They depend on refusing to confuse technical achievement with a mandate to rule.
Reading Star Trek does not confer a right to rule Earth. Building a rocket does not establish competence in constitutional government. Forecasting a technological future does not grant ownership of it.