Mustafa Suleyman helped build two of the most consequential AI companies of the last decade. When he says the wave is coming and we are not prepared, he is worth listening to.

There is a useful corrective buried in the history of innovation diffusion. When Everett Rogers first mapped how new technologies spread through societies, he labelled those who resisted adoption "laggards" — a word that carried its own verdict. Slower, less intelligent, on the wrong side of history. By the fifth edition of his Diffusion of Innovations, published in 2003, Rogers had revised that judgment. Some of those laggards, it turned out, had refused hybrid corn not out of ignorance but out of accurate foresight: they understood the risks of monoculture, the dangers of supplier dependency, the long-term costs of what was being presented as unambiguous progress. Others who tried to return to conventional seed were sued into ruin for intellectual property violations — for having traces of hybrid corn in their soil. Rogers' final conclusion was precise and uncomfortable: "The consequences of innovation diffusion usually widen the socioeconomic gap between those with higher and lower statuses." This is the frame in which The Coming Wave should be read — and the standard against which it falls short.

Suleyman, a co-founder of DeepMind and Inflection, is not an outside observer of the AI wave. He is one of its makers. That position gives the book something unusual: an insider's account of what is actually being built, written by someone willing to say publicly that the risks are real and the industry's self-regulation is insufficient. That willingness deserves acknowledgment. These are not comfortable arguments to make from inside the technology sector, and the book's existence — its refusal to produce another uncomplicated celebration of progress — represents a form of intellectual honesty that is rarer than it should be. The problem is not the diagnosis. The wave is real, and their account of it is serious. The problem is what happens when they turn from describing the wave to proposing what to do about it.

The response they offer rests on a concept they call containment: a coordinated effort by governments, technology companies, and civil society to manage the pace and direction of development. The word is doing more work than they seem to realise. Containment carries the full weight of Cold War strategic doctrine — the Kennan framework, the logic of American dominance, the architecture of confrontation. As a rallying concept for global cooperation across diverse political systems and competing interests, it is close to self-defeating. It frames the challenge as adversarial and unidirectional, as if AI were a rising ocean rather than a set of decisions being made by specific people in specific institutions for specific reasons. Those reasons — profit, competitive pressure, the structural logic of venture capital — are acknowledged in the book but never seriously interrogated. Containment, it turns out, is a concept that lets the authors gesture at the problem of power without engaging with it directly.

The ten-point framework offered at the book's close — safety first, guardrails, audits, choke points, proactive measures — reads like a responsible policy checklist that arrives without its substance. Each point names a category of action without examining what that action would require, who would perform it, what the obstacles are, or what the track record of comparable efforts has been. Saying that audits matter is not the same as asking who conducts them, what their mandate covers, and what happens when their findings are inconvenient. The authors place considerable faith in the cooperative goodwill of diverse actors — states, companies, civil society — in a landscape they have themselves described as driven by competition, ambition, and profit. The gap between those two positions is never closed.

The deepest contradiction sits closer to the centre of the book. Suleyman introduces the concept of ACI — Artificially Capable Intelligence — as a near-term development: an AI system capable of taking a modest investment and generating substantial returns through autonomous market research, product development, and human oversight management. He presents this as an illustration of what is coming. What he appears not to notice is that it is a precise description of Philip K. Dick's Autofac — a 1955 story about automated factories that, having lost contact with their human supervisors, continue producing according to their own logic, optimising for metrics no longer connected to human need, devastating the landscape in the process. Dick wrote it as a cautionary fable. Suleyman presents the same architecture as an exciting development. The authors acknowledge, elsewhere in the book, that technology's creators routinely lose control of their inventions once released. The ACI vision sits in direct tension with that acknowledgment, unexamined.

What Rogers understood, by the end of his career, was that the optimism built into the language of innovation — progress, adoption, diffusion — was not neutral. It carried a set of assumptions about who benefits, on what timeline, and at whose expense. The displaced factory workers who do not appear in the robotics examples, the white-collar workers whose replacement is described as cost-efficient rather than as a social rupture, the political consequences of technological disruption that continue to fuel reactionary movements across democratic societies — these are not edge cases or temporary friction. They are the wave, experienced from a different position than the one the book is written from.

A more rigorous engagement with the problem Suleyman and Bhaskar have correctly identified would require what Rogers eventually managed: a willingness to take the laggards seriously, to examine the frameworks being used before applying them, and to sit with the discomfort of not having a clean solution. The European Union's AI Act and UNESCO's ethics frameworks represent early attempts at that kind of structured engagement. They are imperfect and contested. So is every serious attempt to engage with a problem this difficult. The book's limitations do not diminish the problem it describes. They are simply an illustration of how far the thinking still needs to go.