The Children of the Helpful State

Picture Kaliya and Soën at sixteen.

Not in a dystopian bunker. Not under soldiers. In a bright European apartment, on an ordinary afternoon, with the same small impatience that belongs to every generation. They want to read something, watch something, make something, say something. The screen asks them to prove their age.

They tap a wallet. A green circle appears: **old enough**.

The screen does not show their birthday. The system says it does not need to. It needs only the answer to one question.

This is the humane version. It is private by design. It is safer than handing a passport to every website. It is meant to keep children away from pornography, gambling, alcohol, manipulation, and the industrial machinery that has discovered how profitably a young nervous system can be kept afraid and awake.

Now imagine the next screen.

The article they want to read is not illegal. Nobody has banned it. Nobody has sent a policeman to their home. But the platform has classified it as politically sensitive, psychologically risky, or likely to expose a minor to material that a compliance system cannot confidently distinguish from harm. The platform asks for another attestation. The attestation is still anonymous. The gate is still described as protection.

Kaliya asks why a reader has to prove adulthood to read an argument.

Soën asks who gets to define the risk.

The answer comes back in the softest voice in the language: **we are doing this for you**.

This scene is fiction. No current EU rule says that every political article must be age-gated, and the European Digital Identity Wallet is described by the Commission as optional. The point is not to report a future law as if it existed. The point is to make visible a structural possibility: once identity, age, risk classification, and access controls are made interoperable, a protective tool can become the default grammar of ordinary life. A permission that was introduced for an exceptional danger can become the way a person enters the public world.

That is the danger worth naming.

## The sentence Lewis actually wrote

C. S. Lewis gives the danger its most unsettling formulation:

> “Of all tyrannies a tyranny sincerely exercised for the good of its victims may be the most oppressive.”

The sentence comes from **“The Humanitarian Theory of Punishment,”** an essay first published in 1949 and later reprinted in *God in the Dock: Essays on Theology and Ethics* (1970). It was not originally a commentary on Brussels, digital identity, or children’s online safety. Lewis was arguing about punishment: what happens when a society stops treating a wrongdoer as responsible for a wrong and starts treating him as a patient to be corrected by experts.

That context matters. The quotation is not a magic password that proves the European Union is tyrannical. It is a warning about a particular mutation of benevolence: when the person being protected loses the right to define the good being imposed on him, and when the authority that claims to cure him acquires no clear stopping point.

The EU is not an accomplished tyranny. It has elections, courts, a Parliament, national governments, public consultations, rights instruments, regulators whose decisions can be challenged, and citizens who can organize against its policies. Its digital laws contain real protections. The Commission’s age-verification design aims to prove an age threshold without revealing a person’s exact age or identity. The wallet framework includes selective disclosure, user control, and the possibility of pseudonyms where the law does not require legal identification.

Those are not cosmetic facts. They are safeguards.

They are also not guarantees that safeguards will survive contact with incentives.

## The protective architecture is already here

The Digital Services Act has applied generally since 17 February 2024. For very large online platforms and search engines — services with more than 45 million monthly users in the EU — it adds duties to identify and mitigate systemic risks. The Commission names risks to fundamental rights, media freedom, public health, children, mental wellbeing, electoral processes, and public security. Large platforms must assess those risks, adapt their systems, submit to audits, share data with regulators and vetted researchers, and offer at least one recommender option that is not based on profiling.

The intention is intelligible. A platform that can reach hundreds of millions of people is not merely a private noticeboard. Its ranking systems can alter what people see, what they fear, what they buy, and what they believe other people believe. Requiring it to understand and mitigate systemic harm is not paternalism by definition. It can be the minimum price of operating a machine at continental scale.

But “systemic risk” is a powerful category. It moves the center of gravity from a concrete illegal act to a predicted effect on a population. Once the object of regulation becomes the risk a system might create, the system can be redesigned around the regulator’s model of a healthy society. A platform does not need to be ordered to ban an opinion. It may decide that the opinion is too costly to recommend, too uncertain to distribute, or too risky to show to the wrong class of user.

The AI Act adds another layer. As of 14 July 2026, prohibited practices and AI-literacy obligations have applied since February 2025, governance rules and obligations for general-purpose AI models since August 2025, and the Act’s transparency rules are due to apply on 2 August 2026. The law is a risk-based framework, not a general ban on thought. It is trying to make systems safer, more legible, and more accountable.

Yet risk-based governance has an internal temptation. The more capable the system, the more categories of human life become legible as risk variables. A child becomes an age attribute. A reader becomes a profile. A conversation becomes a safety classification. A model’s answer becomes a regulated output. The person is not formally reduced to the variable. The infrastructure merely begins to behave as if the variable were the most important thing about the person.

In April 2026, the Commission recommended that Member States make an EU age-verification solution available by the end of the year. The proposed approach can be deployed as a standalone app or integrated into a European Digital Identity Wallet. The Commission says the tool can confirm that someone is above a threshold without revealing exact age, identity, or other personal details. The EU’s identity framework itself requires Member States to provide at least one wallet by the end of 2026, while the Commission describes its use as optional.

Again: this is not a universal identity mandate. It is not proof that Europe has abolished anonymity. It is a set of real, current building blocks, with privacy claims and legal limits, from which future defaults will be assembled.

The question is not whether the first version is benevolent. The question is what happens when every benevolent exception finds a new use.

## The state that never feels cruel

The old tyrant was easy to recognize. He shouted. He censored. He arrived with a list of enemies and a visible force to enforce it.

The new paternalism can arrive as a sequence of successful product decisions.

First, the state asks a platform to protect minors.

Then the platform asks a user to prove an attribute.

Then a wallet makes the proof cheap, interoperable, and familiar.

Then a new risk is identified: misinformation, extremism, self-harm, unhealthy political content, manipulative intimacy, destabilizing images, dangerous advice. Each risk is real somewhere. Each deserves serious work. Each can justify a narrow intervention. But the infrastructure does not know how to remain narrow by itself. It prefers reusable rails.

The result may be a society in which no single law abolishes freedom, yet the cost of exercising it rises at every doorway. The citizen is still free to speak — after classification, authentication, age proof, platform review, recommender demotion, and the quiet possibility that an institution has decided the speech is not good for the citizen who chose it.

The cruelty, if it comes, will not necessarily feel like cruelty. It will feel like friction removed for everyone’s safety.

This is why Lewis’s warning is sharper than the ordinary accusation of censorship. Censorship says: **you may not say this**. Paternalism says: **you should not have to encounter this, and we know better than you what will protect you**. The second can be more difficult to resist because it recruits the conscience of the person being governed. To object is to appear indifferent to children, victims, public health, truth, or safety.

The moral case for protection is real. So is the moral case against making protection the only language in which a citizen can justify freedom.

## What Kaliya and Soën should inherit

Kaliya and Soën do not need a Europe that abandons children to platforms. They do not need a romantic internet where every danger is called freedom and every intervention is called tyranny. They need adults capable of holding two truths at once.

The first truth is that online systems can harm children and adults at scale. Governments have a legitimate duty to constrain exploitation, fraud, coercion, and systems that externalize their damage onto the vulnerable.

The second truth is that a person is not made safer by being permanently translated into a risk score. Protection that cannot be refused, questioned, appealed, or escaped becomes a form of custody. A wallet that is optional today can become practically necessary tomorrow if schools, employers, platforms, banks, and public services all discover reasons to prefer it. A recommendation that is privacy-preserving in its technical design can still participate in a social world where access increasingly depends on presenting the right attribute to the right gate.

The test is therefore not only: **does this tool protect?**

It is also:

- Can Kaliya say no without losing ordinary participation?

- Can Soën challenge the classification without needing permission from the classifier?

- Is there a humanly intelligible appeal when a system gets the person wrong?

- Does the rule expire, or does every successful intervention become permanent infrastructure?

- Can a citizen remain a citizen when the system cannot confidently decide what kind of risk they are?

Those questions do not accuse the EU of tyranny. They ask whether a democratic project can keep its protective hand from becoming a hand that never lets go.

The children in the opening scene are not waiting for a dictator. They are growing up inside a succession of reasonable decisions. Every decision may be defensible in isolation. The shock appears only when they look back and discover that the world has been made conditional without anyone announcing the change.

The EU’s present digital architecture is a documented experiment in governing risk at continental scale. It contains serious rights protections, and it may prevent serious harms. It also creates the possibility of a softer authority: one that does not need to hate its subjects, because it sincerely believes that unrestricted agency is an avoidable source of injury.

That is the world Kaliya and Soën may inherit. Not a completed tyranny. Something more ordinary, and therefore more difficult: a public order so helpful that refusing its help begins to look irresponsible.

Lewis’s warning is not that every person who protects another is a tyrant. It is that good intentions do not supply a stopping point. The stopping point must be built into the rights, the appeals, the alternatives, the sunset clauses, and the courage to leave some people alone.

If Europe can do that, its protections will remain democratic protections.

If it cannot, the first sign will not be a prison door.

It will be a child asking why the world needs to know who they are before it will let them read.

## Sources and attribution

- C. S. Lewis, “The Humanitarian Theory of Punishment,” first published in *20th Century* (1949), reprinted in Walter Hooper (ed.), *God in the Dock: Essays on Theology and Ethics* (1970). The quoted sentence is from that essay, not from *The Abolition of Man*.

- European Commission, [AI Act: application timeline and current implementation](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai), checked 14 July 2026.

- European Commission, [DSA: very large online platforms and search engines](https://digital-strategy.ec.europa.eu/en/policies/dsa-vlops), updated 19 May 2026.

- European Commission, [common approach for EU-wide age-verification technologies](https://digital-strategy.ec.europa.eu/en/library/commission-sets-out-common-approach-eu-wide-age-verification-technologies), 29 April 2026.

- European Commission, [European Digital Identity Regulation](https://digital-strategy.ec.europa.eu/en/policies/eudi-regulation), updated 22 June 2026.

- EUR-Lex, [Regulation (EU) 2024/1183 establishing the European Digital Identity Framework](https://eur-lex.europa.eu/eli/reg/2024/1183/oj).