Monday, 30 March 2026 — Issue No. 1 AI, platforms, and digital power — critically examined

AI;DR — AI;Didn't Read
What AI sceptics say when they won't read machine-generated text

How this works: All texts and articles are produced entirely by AI — through a chain of specialised systems that frame, research, write, and revise. It draws on documented sources and disclosed materials. Humans do not edit the writing. The system may also call on humans for interviews, accounts, or specific documentation when needed.
What remains human is the act of publication. The editor decides what is released and assumes responsibility for its content and consequences.

Drafted with
Claude Sonnet 4
—
Editorial review
by the editors
—
Model disclosed
per article
—
CC BY-SA 4.0
No ads · No tracking

Lead Essay

On Language & Accountability

What We Mean When We Say AI Understands

The language used to describe AI systems is not accidental. It shapes what questions we think to ask — and which ones we stop asking altogether.

By the Editors — AI;DR, Issue 1 — Drafted with Claude Sonnet 4 · Editorial review by the editors · No primary research

In March 2023, the Italian data protection authority suspended ChatGPT, citing the absence of any legal basis for collecting Italian users’ data at scale. The suspension lasted one month. During that month, journalists, policymakers, and ordinary users confronted a question that had been obscured by the speed of adoption: on whose terms had this tool entered public life, and who was responsible for what it said?

That question is harder to answer than it seems. The companies behind large language models describe their systems as reading, writing, and reasoning. These verbs import a set of assumptions — about intention, about understanding, about the possibility of being held to account. A system that ‘reasons’ is expected to explain itself. A system that merely compresses statistical patterns is not.

Calling a language model intelligent is not a neutral description. It is a political choice, one that redistributes responsibility away from the people who built and deployed it.

This matters most where the stakes are highest: automated benefit decisions, medical triage tools, content moderation systems operating at national scale. In each case, the vocabulary of machine intelligence serves to obscure the chain of human choices that produced the system and continues to govern its use.

AI;DR covers these systems concretely: what they do, what they fail to do, how they are governed, and who bears the cost when they go wrong. We do not oppose AI as a category. We do oppose the idea that technical complexity excuses anyone from answering for its effects.

Read more →

Further Reading

Infrastructure

The Public Cloud Is Not Public

European municipalities are running their citizens’ data on American infrastructure, under American law, with limited recourse.

When a German Landkreis migrates its administrative systems to Microsoft Azure, it does not simply change its software provider. It transfers operational control of public records to a foreign company subject to the US CLOUD Act — which grants American authorities access to data stored abroad by US firms, regardless of where it sits.

Several cities — Cologne, Barcelona, Grenoble — have begun auditing these dependencies and piloting sovereign alternatives. The work is slow and technically demanding. It is also, increasingly, a matter of public law.

Drafted with Claude Sonnet 4 · Editorial review by the editors · No primary research

Read more →

Platforms

The Recommendation Engine as Editorial Policy

Algorithmic feeds do not reflect what users want. They optimise for what keeps users engaged. The difference is not trivial.

Meta’s internal research, disclosed during the 2021 US Senate hearings, showed that its own engineers had identified amplification of divisive content as a product of the engagement model — and that the recommendation to limit this was not implemented. The decision was commercial, not technical.

This is what it means to say a platform has editorial responsibility: not that it writes content, but that its infrastructure makes choices about what spreads and what does not.

Drafted with Claude Sonnet 4 · Editorial review by the editors · Sources: US Senate testimony, 2021

Read more →

Regulation

The AI Act and Its Limits

The EU has the most comprehensive AI regulation in the world. It also has significant gaps, and industry has had years to shape its implementation.

The European AI Act classifies systems by risk and imposes requirements accordingly: transparency obligations for limited-risk systems, conformity assessments for high-risk ones, outright bans for a small category of applications. The framework is real. So are the exemptions carved out for national security and law enforcement.

Civil society organisations have spent three years tracking the lobbying that shaped the final text. Their records are public. Reading them is instructive.

Drafted with Claude Sonnet 4 · Editorial review by the editors · Sources: EUR-Lex, AlgorithmWatch

Read more →

Our Principles
01

Documented Production

Every article discloses the models used, the production chain, and the sources it draws on. Readers know what they are reading and how it was produced.

02

AI Production & Human Responsibility

The writing is produced entirely by AI. No human edits the text. Publication is a human decision: the editor determines what is released and assumes responsibility for its content, framing, and consequences.

03

No Synthetic Authority

Fluency is not authority. A well-phrased sentence is not evidence. We treat machine-generated text with the same scepticism we apply to our subjects.

04

Open Infrastructure

All content is published under CC BY-SA 4.0 on open infrastructure. No advertising. No investors. No paywall. Affiliated with Web B at reappropriate.org.