David Harvey: The Word that Swallowed Everything

This could be seen as a companion piece to “The Art of Not Deciding” which I published on 29 April 2026. That article dealt with decision paralysis. One aspect of that is fear and accompanying that is a desire to stay safe – from criticism, condemnation. Better to do nothing than do something and be criticised for it.

In this article I look at the issue of safety and the word “safe” and how that word has been weaponised.

Let me begin with a word. Not a law, not a regulation, not even a policy — just a word.

“Safe.”

It is four letters long and was once perfectly serviceable as a descriptor. The ship was safe in harbour. The children arrived home safe. It told you something had been accomplished: a journey through uncertainty had ended without catastrophe.

What has happened to that word since is one of the more instructive exercises in linguistic capture I have observed in the course of a long engagement with the law and its surrounding culture.

The word has escaped its descriptive function entirely and reinvented itself as a governing principle. It no longer tells you how things turned out. It tells you how things must be arranged in advance.

This is not a trivial observation. Words that become governing principles carry enormous practical power. They set the terms of argument. They determine who bears the burden of justification. And “safety,” deployed as a governing principle, carries a burden of justification so asymmetric as to be almost irresistible: whoever invokes it claims the moral high ground before the discussion has begun, and whoever questions it must answer for the harm they appear to be endorsing.

That is not reasoning. That is rhetorical pre-emption. And we have been living under it, in various degrees of comfort, for the better part of three decades.

I have watched this from a number of vantage points – some professional, others more general. The pattern is consistent. A perceived harm is identified. “Safety” is invoked as the response. A regulatory or supervisory framework is proposed. And the question of what is lost — what autonomy, what capacity, what human development is foreclosed by the intervention — is treated as, at best, a secondary concern, and at worst, evidence of bad faith on the part of those who raise it. That sequencing is not accidental. It is the operational logic of safety culture.

The precautionary mind

The proper discipline for managing risk is not new. Actuaries have practised it for centuries. Judges apply versions of it daily in the allocation of legal liability. Its core insight is simple and not at all alarming: risk cannot be abolished; it can only be identified, assessed, and distributed.

Every decision to mitigate one hazard carries a cost — direct financial cost, opportunity cost, and the cost of the risks that mitigation itself creates. A medication that eliminates one disease may cause another. A road barrier that prevents one type of accident may cause a different one. A playground drained of all challenge produces children who have not learned to fall.

Classical risk management asks the question that contemporary safety culture has largely ceased to ask: compared to what? What is the probability of harm? What is its severity? What is the cost of preventing it — in all senses of “cost”? And what beneficial outcomes are precluded if prevention is pursued at the expense of everything else?

These are not hostile questions. They are the questions of any adult engaging seriously with an uncertain world. The precautionary principle, by contrast, does not engage with them. It instructs that where any conceivable harm exists, action should be taken to prevent it, regardless of probability, severity, or cost. That is not risk management. It is the abdication of risk management dressed in the costume of prudence.

“Safety, properly understood, is the outcome of good judgement, not a substitute for it. When it displaces judgement, it does not make us safer. It makes us supervised.”

The practical consequence is a regulatory architecture that expands in one direction only. Each new safety measure creates new bureaucratic infrastructure, new professional interests invested in the perpetuation of that infrastructure, and new categories of formerly acceptable behaviour rendered non-compliant.

I have written before about the incremental nature of this expansion in the context of internet content regulation — the way in which the same agencies recur, the same language recycles, and the same objective advances under successive banners.

The same logic operates across virtually every domain of public and private life. Health and safety regulation. Safeguarding frameworks. Due diligence requirements. Risk assessments mandated for activities that were once managed by the application of ordinary sense. The individual layer is defensible; the cumulative architecture is a transformation in the relationship between the citizen and the state that nobody put to a vote.

What we are doing to our children

The most consequential site of this transformation is the upbringing of children, and it deserves careful attention. The case for protecting children from risk has obvious moral force. Children do not possess fully developed risk assessment capacities, and adults have both a legal and moral duty of care.

Nobody of good sense disputes this. What has changed is the threshold at which protection is invoked, the category of hazards against which it is deployed, and — most significantly — the implicit theory of child development that underlies it.

Developmental psychology, which has had rather longer to observe children than the modern regulatory state, has long understood that competence is acquired through attempt and occasional failure.

The bumped knee teaches bodily limits. The lost game teaches losing, which is to say it teaches equanimity and recovery. The social miscalculation teaches repair. These experiences are not unfortunate side-effects of growing up. They are its mechanism.

The child who is shielded from all such encounters does not emerge unscathed; they emerge unequipped. The curriculum of childhood cannot be outsourced to a safety audit, because its content is precisely the encounter with difficulty that safety audits are designed to prevent.

We have, to put it plainly, confused protecting children from harm with protecting them from experience.

They are not the same thing. Harm is a wound; experience is education. A culture that cannot distinguish between them will produce — is, I would argue, already producing — young people who have not been equipped with the internal resources necessary to navigate an adversity that the world will visit upon them regardless of whatever protections were arranged in childhood.

The rates of anxiety, fragility, and difficulty tolerating uncertainty now observable among young people in many Western countries have emerged precisely during the period in which risk-averse parenting and educational culture became dominant.

That correlation does not establish causation; but the hypothesis that over-protection contributes to the problem it professes to prevent is taken seriously by serious researchers, and it should be taken seriously by everyone else.

“A generation raised in a padded environment does not emerge toughened. It emerges unaccustomed to hardness — and not particularly well-placed to deal with a world that remains, stubbornly, hard.”

Risk, agency, and the civic capacity for self-governance

Risk tolerance is not merely a personal psychological trait, of interest only to those in the business of individual wellbeing. It is a civic and economic capacity, and its atrophy has consequences that extend well beyond the individual.

Entrepreneurship requires accepting the possibility of failure.

Scientific enquiry requires the willingness to be wrong, publicly.

Political dissent requires the acceptance of social cost.

Artistic innovation requires the production of work that may be rejected, ridiculed, or ignored.

None of these activities are available to people who have been conditioned to regard risk as a pathological condition requiring professional management.

A culture that systematically undermines risk tolerance does not simply produce more cautious individuals; it erodes the social substrate on which intellectual and economic dynamism depend.

There are economic costs to this that seldom surface in safety discourse, partly because they are invisible — they appear in the ledger of things not done, businesses not started, drugs not developed, procedures not attempted. The regulatory compliance costs that deter the small enterprise. The liability exposure that deters medical innovation. The reputational risks that deter the public intellectual from taking an unpopular position.

These costs are real. They simply do not appear on any risk register, because risk registers are designed to count the harms that might be caused by action, not the benefits lost through inaction.

The precautionary principle has no accounting mechanism for what it forecloses. That is a significant deficiency in any instrument presented as a tool of rational governance.

The political utility of safety language

Here I come to what I consider to be the central issue — one which, as readers of my writings will know, and which I have approached from a number of directions over the years.

Safety, as a political concept, has properties that make it peculiarly useful to those who wish to extend the supervision of human behaviour without incurring the democratic resistance that explicit claims to control would attract.

It is difficult to oppose. Few will argue openly against protection from harm. It is infinitely expandable: the category of what constitutes a hazard can be widened without logical limit — from physical injury to emotional discomfort, from offensive speech to dietary choices, from personal conduct to the content of one’s online expression.

It is transferable across ideological territory: governments of the left invoke it to regulate speech and protect designated groups from distress; governments of the right invoke it to regulate personal behaviour and enforce social norms. Both deploy it sincerely. Both benefit from the authority it confers.

The result is a bipartisan expansion of supervision in which disagreements about whose safety matters obscure a shared and unexamined assumption that safety is the appropriate standard by which all behaviour should ultimately be judged.

I have observed this mechanism operate in the context of internet content regulation with some precision. The same language recycles — “safety, harm, trust, resilience” — under successive initiatives, from successive agencies, toward a consistent objective. The Christchurch Call. The Safer Online Services proposals. The Media Reform agenda. The Education and Workforce Committee inquiry. The Department of Internal Affairs, which one might have thought had learned a lesson from the deep-sixing of its previous overreach, reconstituting itself as the vehicle for the next attempt.

Each initiative presents itself as a response to a specific, identified harm. Cumulatively they constitute a long, determined, and substantially consistent project: the regulation of content, expression, and digital behaviour under the banner of safety. The banner does not describe the project. It protects it from scrutiny.

“Safety, mobilised as a political concept, is power that declines to name itself as such — and is all the more effective for that refusal.”

A framing device that pre-empts rational analysis is useful to anyone seeking to advance a policy agenda that might not survive rational analysis. The safety frame instructs us that the question is not whether a proposed intervention represents an appropriate exercise of state power, but whether we are in favour of harm or against it. It converts a question of governance into a moral binary.

That conversion is not accidental, and its consequences are not benign.

The managed life

Taken together, these tendencies sketch the outline of a model of human life that is, in its deepest structure, administrative. The good life, in this model, is the managed life: lived under the supervision of credentialled professionals, insured against foreseeable adversity, compliant with applicable frameworks, and protected from experiences whose value has not been officially assessed and approved.

The ideal subject of this world is not an autonomous person — self-directing, risk-accepting, accountable for the consequences of their own choices — but a managed participant: grateful for protection, deferential to expertise, and disposed to understand their own limitations primarily in terms of the hazards their unsupervised choices might generate.

I want to be precise about what I am not saying here. I am not suggesting that this model is the product of conscious malice. Most of the individuals who design and implement safety frameworks believe in them sincerely, often in direct response to genuine harms that required genuine attention.

The doctor who recommends the safest course, the regulator who closes the loophole, the official who mandates the risk assessment — none of these people is a villain.

They are operating within a cultural framework that treats safety as the paramount value, responding to the incentive structure that framework creates. The road to supervision, like most important roads, is paved with benign intentions.

But the cumulative effect of individually defensible decisions, operating within that framework, is a systematic reduction in the space available for autonomous human action. And that reduction, once achieved, is not easily reversed.

Bureaucratic infrastructure, once established, develops its own interests, its own vocabulary, its own account of why it is necessary. The apparatus that was created to address a specific harm acquires the capacity to identify new ones. The officials who were appointed to manage a defined risk develop an institutional interest in finding additional risks to manage. This is not conspiracy; it is the ordinary operation of institutional self-preservation. But its effect is the steady colonisation of more and more of human life by the supervisory state.

What is forfeited

What is lost in this process does not appear on any official ledger. It cannot be entered in a risk register or quantified in an impact assessment. What is lost is something that does not yield to those instruments: the quality of a life lived with genuine agency.

To accept risk is to accept responsibility for the outcome — which is to say, to be genuinely the author of one’s own life rather than a managed participant in someone else’s assessment of it.

The satisfactions of difficulty, of self-reliance, of achievement wrested from resistance — these are not available to the supervised life. They require exposure to the possibility of failure, which the safety framework is designed to eliminate.

There is also a civic loss which I regard as serious. The capacity for democratic self-governance requires citizens who believe themselves capable of exercising judgement, tolerating genuine disagreement, and accepting the uncertainty that inheres in the collective management of contested questions.

A culture that systematically trains its members to regard themselves as incapable of assessing their own risk will produce, over time, citizens who are correspondingly less capable of doing so — who look to authority for guidance on what is permitted, who are less willing to challenge official consensus, and who have been habituated to experience uncertainty not as the ordinary condition of a free life, but as a problem requiring institutional resolution.

You cannot produce autonomous democratic citizens by raising and educating people to be managed subjects. The two projects are in direct tension.

Let me be clear about what this argument does and does not require. It does not require the view that safety considerations have no place in public life. The abolition of child labour was an achievement. The seatbelt was an improvement. The food safety standard addressed a genuine harm. These represent real contributions to human welfare, and they were not produced simply by leaving everything to individual preference. ‘

The question is not whether safety matters but whether it should function as an absolute trump — a value so paramount that it pre-empts every other consideration and provides sufficient justification for any intervention advanced under its banner.

The answer to that question, I would suggest, is plainly no. A mature polity would hold safety as one consideration among several, weighed against autonomy, against dynamism, against the developmental requirements of the young, and against the straightforward human interest in living a life of one’s own construction rather than one curated by the risk-averse on one’s behalf.

It would ask not merely “what is the safest course?” but “what kind of people do we become if we always take it?” And it would recognise — as it too rarely does — that a word which began as a description of an outcome has been conscripted as an instrument of governance. That conscription has not been innocent, whatever the intentions that attended it.

The recovery of a vocabulary adequate to this situation requires, at minimum, the willingness to ask the questions that the safety frame is designed to prevent: compared to what? At what cost? To whose benefit? And who, precisely, decided that this was our problem to manage rather than our risk to accept?

These are not dangerous questions. They are the questions of a free people. The fact that they have become difficult to ask in public without appearing to endorse the harm one is questioning is itself, I would submit, a measure of how far the capture has proceeded.

David Harvey is a former District Court Judge and Mastermind champion, as well as an award winning writer who blogs at the substack site A Halflings View - Where this article was sourced