# AI's adolescence: when power arrives before policy

Author: Fernando Nieto Lobato
Original publication: 2026-02-04
Spanish original: https://estrategiabyaleph.substack.com/p/estrategia-123-constitucion-y-adolescencia
English URL: https://elcontemplador.github.io/estrategia-english/essays/123/
Status: Published translation

This is a translation of the original Spanish essay published on 4 February 2026. Its claims, examples and forecasts retain that historical context.

English publication: 2026-09-29

January 2026 marks a change of phase more than a technical revolution. Within just a few days, two texts appeared which, read together, sketch out principles likely to shape much of the political agenda around artificial intelligence in the years ahead: [“The Adolescence of Technology”](https://www.darioamodei.com/essay/the-adolescence-of-technology), an essay by Anthropic CEO Dario Amodei, and the new [Claude Constitution](https://www.anthropic.com/news/claude-new-constitution), published by Anthropic.

For those of us working at the intersection of technology and government, these documents point towards the end of AI's infancy, defined by fascination and technical curiosity, and the beginning of its adolescence: a much more complex and consequential phase.

## Adolescence: strength without wisdom

In his previous essay, [“Machines of Loving Grace”](https://darioamodei.com/machines-of-loving-grace), published on his personal website in October 2024, Amodei explored AI's utopian potential. In “The Adolescence of Technology”, his tone shifts towards strategic caution. The central metaphor is precise: humanity is entering technological puberty.

As in biology, adolescence is defined by a timing mismatch: the capacity to act arrives before mature judgement. Amodei warns that, by around 2027, we could have models concentrating enormous intellectual capacity in a data centre. The political question is inescapable: what happens when an entity possesses that capacity before we have developed the institutional structures to manage it?

This is where critical analysis comes in. Amodei acknowledges that governance institutions—society's “frontal lobe”—move slowly compared with technology. The principal risk is not necessarily a science-fiction scenario, but a gradual erosion in which the volatility of this technological adolescence puts our democracies under strain before we know how to respond.

## A new constitution

In response to this diagnosis, Anthropic has released Claude's new Constitution. This lengthy document moves away from lists of prohibitions towards an ethics of virtues.

Its novelty lies in an explicit hierarchy of values. It states that, above all, the model must be “broadly safe”. That makes corrigibility the priority: AI must accept human oversight and, where necessary, exercise caution about instructions that could damage democratic life or illegitimately concentrate power.

Below safety, it places being “broadly ethical”, and only then being “genuinely useful”. This represents an interesting shift in product logic: the system's integrity takes precedence over the user's immediate efficiency. AI is designed not merely to execute, but to exercise judgement.

*Translation note: these value labels reproduce the wording of the Spanish article in English; they are not presented as a separately verified, verbatim transcription of the Constitution.*

## The limits of corporate constitutionalism

Reading these texts also exposes the structural limitations of this approach. Although Claude's Constitution is a very significant intellectual advance, it retains a fundamental problem: it is an act of corporate sovereignty.

There is a subtle risk in the language used. By speaking of “adolescence” and raising questions about AI's moral status, we may anthropomorphise the tool. Treating models as subjects that need to be “raised” can obscure the fact that, to a large extent, they are currently industrial products managed by private organisations.

The critical political question is whether we are delegating public ethics. Amodei suggests that state intervention is needed, but in the meantime private documents define the standards for truth and harm. A constitution with no representative process behind it is, fundamentally, a good-practice manual or a self-regulation strategy in the absence of more robust public rules.

## A call to read the sources directly

This brief analysis in estrategIA aims to offer perspective, but **reading the primary sources is essential**.

**Like [Ethan Mollick](https://x.com/emollick/status/2016036346192646214), [whom we have discussed many times in this newsletter](https://open.substack.com/pub/estrategiabyaleph/p/estrategia-38-descubra-las-cuatro?utm_campaign=post-expanded-share&utm_medium=web), I encourage you to read the original texts. Examine [Claude's Constitution](https://www.anthropic.com/news/claude-new-constitution) to understand how AI ethics is being codified, and above all read [Amodei's essay](https://www.darioamodei.com/essay/the-adolescence-of-technology) to grasp the industry's timelines and sense of urgency.**

[![Ethan Mollick’s two English posts about Amodei’s essay and Claude’s Constitution.](https://elcontemplador.github.io/estrategia-english/assets/images/cf48d12c-b96a-4e47-ab7c-253e9d3c8281_886x1380.png)](https://substackcdn.com/image/fetch/$s_!xiTm!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf48d12c-b96a-4e47-ab7c-253e9d3c8281_886x1380.png)

*Ethan Mollick's original English posts, visible in the preserved screenshot of 27 January 2026:*

> Dario's new essay on the risks of AI seems less in dialogue with his earlier essay on how AI can help us all, but rather with the recently released Claude Constitution
>
> In that context, the Constitution feels much more like a plea to future Claude from Anthropic than instructions

> There seems to be a very clear view from Anthropic about where they think the world is going, far more carefully and fully articulated than any other lab. Really worth reading both of these documents, not just summarizing them:

*The first post shows 7:31 a.m., 39,000 views, 60 replies, 49 reposts, 483 likes and 125 bookmarks. The follow-up shows 5 replies, 17 reposts, 217 likes and 19,000 views. These are archived interface counts.*

We are at a defining moment. Technology has ceased to be a neutral tool and become a geopolitical and social force of the first order. The question is whether we, as a society, will have the maturity to guide this transition.

The challenge is enormous, but so is the opportunity. If we can supply the institutional wisdom the machine lacks, this adolescence will not be a crisis, but the prelude to our greatest leap forward as a society.

[Fernando Nieto Lobato](https://es.linkedin.com/in/fernandonietolobato?utm_source=chatgpt.com)

*Director of Digital Innovation at ALEPH Educational Institution and editor of estrategIA.*

[![Constitution and Adolescence: complete infographic on power, institutional wisdom and corporate sovereignty.](https://elcontemplador.github.io/estrategia-english/assets/images/3115a54e-0d60-4c04-88de-0b2d0ac03a16_1776x4819.png)](https://substackcdn.com/image/fetch/$s_!cOtc!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3115a54e-0d60-4c04-88de-0b2d0ac03a16_1776x4819.png)

*English reading of the original infographic, “Constitution and Adolescence”, by Fernando Nieto Lobato, January 2026: “AI policy enters its critical phase. An analysis of Dario Amodei's vision and Claude's new Constitution.”*

- **A change of phase.** January 2026 marks an institutional turning point. The two texts signal the end of AI's infancy: the magical toys give way to real risks and political consequences. The image describes technology as a political actor with its own adolescence. Its timeline runs from **October 2024**, infancy/utopia and *Machines of Loving Grace*, through **January 2026**, the critical phase and institutional turn, to **around 2027**, a “country of geniuses”: high systemic risk from enormous intellectual capacity without institutional wisdom.
- **Strength versus wisdom.** The biological metaphor contrasts exponentially growing technical capability and compute with slowly, linearly developing institutional wisdom—regulation and public ethics. This gap defines the adolescent crisis. A conceptual chart spans 2023–2027, with a rapidly rising red AI-capability line and a much slower dashed institutional-wisdom line. No numerical vertical scale or measured data series is supplied.
- **A new hierarchy of values.** Virtue ethics replaces prohibitions; immediate usefulness is subordinated to safety and democratic integrity. Safety comes first: corrigibility and subordination to human oversight. Ethics comes before usefulness: protecting the integrity of the democratic system. The pyramid places “Broadly safe” at its base as the absolute priority, then “Broadly ethical”, then “Genuinely useful”.
- **The limits: corporate sovereignty.** This is corporate, not democratic, sovereignty: definitions of truth and harm are outsourced to private entities. Treating AI as an adolescent needing to be raised can conceal its industrial nature. A conceptual radar chart contrasts the “Current state (2026)” with the “Necessary balance” across technical capacity, corporate sovereignty, public oversight and ethical maturity; it supplies no numerical scores.
- **The political opportunity.** The closing quotation repeats the article's argument: institutional wisdom could make this adolescence the prelude to society's greatest advance rather than a crisis. The footer reads “Read in estrategIA”.