Editorial Posture

AI Affairs is an analytical publication, not an advocacy publication. Where its sister publication, The Critical Infrastructure Review, takes positions and argues for them, AI Affairs takes the questions — and treats the work of asking them clearly as the editorial mission.

We assume the reader uses AI every day. We do not explain what an LLM is. We do not cover model releases as news. We do not run benchmark coverage, prompt lists, or product walkthroughs. We do not chase the news cycle. We publish work designed to be worth reading in five years.

We are skeptical of the four ideologies currently flooding AI commentary: techno-utopian acceleration, safety-doomerism, Luddite refusal, and the careerist neutrality that treats AI as a beat to cover rather than a phenomenon to understand. We share intellectual ground with the analytical tradition that runs from Foreign Affairs and The New York Review of Books through American Affairs, Asterisk, Works in Progress, and Stratechery — publications that take their subjects seriously enough to write slowly about them, with craft, and with a willingness to be wrong in public.

Our five core editorial commitments

Take AI's second-order effects as the central question. First-order coverage (model capabilities, benchmarks, releases) is well-served and uninteresting. Second-order coverage (what AI does to labor markets, to taste, to authority, to attention, to capital flows, to relationships, to the structure of work, to the meaning of expertise) is under-served and consequential. That is the beat.
Range without scattering. The publication ranges across governance, economics, science, work, art, identity, and culture — but every essay reaches the same kind of question: what is this technology doing to the institutions and forms it touches, and what should serious people do about it?
Craft as editorial position. The publication looks, reads, and edits like a magazine that respects its readers' time. Long essays are long because they have to be; short pieces are short because that's enough.
Authority through depth, not speed. We will publish less than the AI commentary aggregators. We will be late on news and on time for arguments that last. The compounding asset is the archive, not the feed.
Independence from the AI industry's PR machine. We will accept sponsorships and institutional subscriptions from AI companies, but we will not accept editorial input, embargoed-but-supportive coverage exchanges, or "thought leadership" placements masquerading as essays. Sponsor relationships are clearly demarcated. Editorial decisions are unilateral.

What we cover

AI's effect on labor markets, professions, and the meaning of expertise
AI's effect on capital allocation, corporate structure, and industrial policy
AI's effect on governance — regulation, antitrust, national security, geopolitics
AI's effect on science — discovery rates, lab structure, the changing role of the researcher
AI's effect on art, taste, attention, craft, and the cultural meaning of authorship
AI's effect on education, expertise, and the credentialing economy
AI's effect on identity, relationships, and the texture of daily life
The history of AI as a technology, an idea, and a cultural object
The institutions building AI — labs, capital, talent, infrastructure — when they meet society

What we do not cover

Model release news, capability benchmarks, evaluation leaderboards
"How an LLM works under the hood" technical explainers
Prompt lists, productivity tips, "best AI tools of [month]" content
Hot takes responding to last week's incident
AI safety theology in either the doom or the dismissal direction
The personalities of the AI principals as celebrities (we cover them as institutional figures only)

On accuracy, corrections, and naming

AI Affairs names institutions, companies, and public figures when the public record supports it. We discuss public statements, publicly reported events, and publicly disclosed numbers as fair editorial comment, and we link to the underlying sources. We aim to characterize subjects fairly even when our analysis is unfavorable.

We do not assert facts beyond what reporting, primary sources, or peer-reviewed work supports. Where we draw inferences from reported facts, we say so. Where the underlying record is contested or partial, we say so.

Subjects who believe we have misrepresented them — or anyone who has identified a factual error — are invited to write to editor@aiaffairs.org. Substantive corrections are issued promptly, dated, and noted on the affected piece. We will not retract pieces for unfavorable analysis we stand behind; we will correct factual mistakes.

This posture applies equally to the AI companies, professional firms, and public figures whose work the publication discusses. We are neither hostile to nor captured by the AI industry. The standard is the same in either direction: the public record is what we report from, and we link to it.