Anthropic’s draft prospectus published on 26 October 2026 offers investors, indeed the whole world, an extraordinary choice. But behind the invitation to trust Anthropic’s leadership of AI’s ‘age of miracles’ lies a much darker reality.
Either we accept an unchecked rush towards ‘catastrophic or existential risks to humanity,’ in Anthropic’s words, or we allow Anthropic and other Big Tech corporations control over whether frontier AI advances, and on what terms—let them make the choices about when economic profit must, occasionally and temporarily, cede to regard for humanity’s other interests.
Viewers of Christopher Nolan’s Odyssey (let’s forget Elon Musk’s sham alternative) may have a sense of déjà vu. Odysseus’ crew also faced a horrific choice between sailing towards the whirlpool of Charybdis—which would kill them all—or steering towards Scylla’s island, from which the monster would certainly kill some of the crew as they passed. Odysseus chose the latter course and sacrificed some of the men, and a bitter crew member was told: ‘that’s what leadership means’.
Well, maybe, in the world of Greek myth. Maybe also in the mythical world of Silicon Valley and Wall Street. But in the wider world, there are different choices to be made about AI.
What Anthropic’s supposedly ‘bold but sane’ leadership conveniently ignores are two crucial points.
We are not in a Greek myth
First, that, while the Charybdis of the latest frontier AI models turning on humanity is a major risks, serious also are the risks of harming the social fabric that flow from intensive and poorly regulated adoption of commercial AI tools. Psychologists have found that intensive ChatGPT use will deskill our children and undermine their desire to learn, creating ‘cognitive debt’. Many, including Big Tech leaders themselves, fear that widespread AI adoption in business will devour jobs, indeed whole sectors of work, substituting for human labour. And a whole range of commentators fear that the arts will be damaged by AI tools like Sora which provide short-cuts to the hard-earned skills on which real creativity is based.
Faced with such evidence of the risks when good actors use ordinary AI—rather than the extreme Charybdis of advanced AI models spiralling out of human control, or bad actors using AI against the world—surely the rational course for society would be to pause, or at least reconsider, AI use? Why not, as the US Brookings Institution put it in an authoritative survey of AI in education this January, move towards ‘AI-resistant choices’? Otherwise, we may end up defenceless against the predatory impact on the social fabric that unconstrained use of commercial AI tools brings.
We must look beyond a bad social contract
Second, Anthropic’s supposed leadership blocks from us from seeing that there really is an alternative approach to AI that avoids the choice between Big Tech’s Scylla and Charybdis. In fact, there is whole ocean of possibilities for using AI for social good and avoiding its social and environmental harms, provided that workers and citizens, not just investors and managers, are allowed to make the choices.
But to grasp this we must see that what is currently on offer from Silicon Valley and most governments across the world is in fact a very bad social contract with AI: a contract that accepts without questioning the AI products that Big Tech is offering us as its route to making profit by recouping the vast costs of developing AI to date.
That bad social contract requires all of us to adopt commercial AI tools as widely and deeply as possible in our lives, whether at work or for leisure. It requires us to adapt our behaviour to suit the AI industry’s profits.
That’s one reason workers in so many sectors are being told they simply must adopt AI, regardless of their own doubts about its effectiveness or usefulness. Another is the fear that, if an organization doesn’t adopt AI enthusiastically, its competitors will leap ahead. The same logic translates to the global scale, where the US’s AI policy is driven by little more than a fear of Chinese economic competition.
But this economics-driven AI race pays no attention to the social costs that will likely flow from adopting AI without regard to the potential harms for society: a world where far fewer people have the chance of meaningful jobs, or even entering the job market; a world where institutions built on respecting human knowledge and expertise are undermined by indiscriminate use of AI tools that redefine expertise in Big Tech corporations’ interests.
Pacing the frontier is not enough
Meanwhile Big Tech’s supposedly responsible leadership of the AI race is riven with contradictions. Anthropic’s CEO Dario Amodei talks of ‘pacing’ AI’s ‘frontier’, without ever suggesting that Anthropic should stop developing the frontier AI models that his prospectus acknowledges could have catastrophic impacts on humanity.
OpenAI hardly offers a sane alternative. The company’s Chief Scientist acknowledged in September the possibility that ‘progress in generalisable alignment may not sufficiently outstrip progress in general model intelligence’: in other words, AI may very soon become just too smart to guide, let alone control. Another OpenAI scientist, Daniel Selsam, put it even more bleakly, noting that AI models may get to the point ‘where they can shape the world unconstrained by human will’, through a process that the industry calls ‘recursive self-improvement’: AI models refining themselves without human involvement.
Little wonder that, after some hesitation, OpenAI pulled the release of its Astra model and now hesitates over its own stock market flotation.
Democracy is based on the principle that societies’ decisions are shaped by human will. When assumptions about AI’s inevitability suggest otherwise, societies must reject Big Tech’s distorted choices and substitute others.
That means seeing that what sustains the supposed whirlpool of ‘inevitable’ AI development is the belief that AI as technology ‘naturally’ moves towards ever greater speed, scale and efficiency. But that’s only true if societies allow technology’s development to be driven by those economic values.
The sea change we desperately need
Instead, we can substitute them with social values—such as maintaining for everyone the opportunity to contribute meaningful to the economy, protecting the integrity of how young people are trained in the capacities they need, and sustaining the social fabric itself. A very different vision of the benefits that AI might bring to society, if carefully used, begins to emerge.
This is the sea change in how societies think about AI that is desperately needed: the change of direction which the seeming calm of Anthropic’s mad prospectus denies.
The Anthropic flotation will no doubt go ahead. Meanwhile workers and citizens everywhere must start insisting on their right to lead the decisions that their workplaces and institutions take about AI: choices about when, and when not, to adopt AI, and on whose terms.
This post gives the views of the author and not the position of the Media@LSE blog, nor of the London School of Economics and Political Science. Nick’s latest book – Predatory AI: How We Can Resist Corporate Capture of the Human Mind (Polity) – is published on 13 November 2026.