Archive
The Archive, By Department.
The full archive of the publication, grouped by standing department. We do not paginate; if a piece has been retired we say so on a dated retirement note. The Cornerstone column carries the publication's definitional pieces.
Cornerstone
3 pieces- 2026-05-28
Model Cards in 2026 — What the Original Paper Got Right and What Gets Ignored
Margaret Mitchell and co-authors published the model-cards paper in 2019. Seven years later, the artefact is the field's most-cited transparency primitive and one of its most-truncated. A reread of…
- 2026-05-10
Anthropic's Mechanistic Interpretability Work at Year Five — What Shipped
A working census of what the lab's mechanistic-interpretability program has actually published since the transformer-circuits thread opened in 2021. Sparse autoencoders, feature dictionaries, attri…
- 2025-10-08
What "Black Box AI" Actually Means in 2026
The phrase has done more rhetorical work in the last three years than almost any other AI term. Most of that work has been imprecise. A working definition, three useful distinctions, and a note on …
Notes
7 pieces- 2026-04-09
NYT v. OpenAI — 20 Million Logs in Discovery
On January 5, 2026, Judge Sidney Stein affirmed a magistrate's order compelling OpenAI to produce twenty million anonymised ChatGPT logs into the New York Times's discovery in the Southern District…
- 2026-03-31
The $1.5 Billion Settlement — What Bartz v. Anthropic Means Going Forward
The largest publicly reported recovery in US copyright history settles a narrow legal question and opens a wider operational one. A working note on the Bartz ruling, the settlement structure, the u…
- 2026-03-18
The Ten Hardest Auditability Problems in Agentic AI
A working list of the genuinely unsolved technical and institutional problems in agentic-system audit. Not a wish list. The actual hard ones, with notes on why each remains unresolved and what woul…
- 2026-02-04
Black Box AI vs. Agentic OS: A Comparative Framing
Two of the most-searched phrases in the AI category, both of them imprecise, frequently confused. A note on what each actually means in 2026, why they are sometimes mistaken for each other, and how…
- 2026-01-21
Why Some Founders Are Choosing Transparency as a Moat
An unusual strategic position is emerging in the agentic category: small operators using auditability not as a regulatory tax but as a competitive lever. We look at the structural reasons it works …
- 2026-01-07
Open vs. Closed in 2026
The open-versus-closed debate has been treated, for the last several years, as a politics question. By 2026 it is a procurement question. A note on what each side has done well, what each side has …
- 2025-10-22
Why Auditability Is the New Differentiator in Agentic Stacks
For most of the AI cycle, the differentiator was capability. By 2026, in the agentic-system category specifically, it has shifted. The firms winning enterprise procurement reviews are the ones whos…
Field Reports
7 pieces- 2026-06-11
The Activation-Patching Renaissance — What Circuits Research Found in 2025
Activation patching, a technique that traces causal contribution of specific model components to specific outputs, moved from a niche mechanistic-interpretability technique to one of the field's mo…
- 2026-06-08
AI Red-Teaming Methods Compared — Anthropic, OpenAI, DeepMind
Three frontier labs publish red-teaming methodology. The methods diverge enough that a reader who reads them carefully can form a working view of what each lab thinks pre-deployment safety evaluati…
- 2026-05-19
DeepMind's Frontier Safety Framework — How 'Critical Capability Levels' Work
Google DeepMind's Frontier Safety Framework defines a tiered evaluation regime for frontier model capabilities and binds the lab's deployment posture to thresholds it has committed to in public. A …
- 2026-02-17
The Compliance Edge: Why AI Marketing Stacks Need Audit Layers
AI marketing was, until recently, an unregulated category. The shift to agentic marketing pipelines — automated outreach, automated segmentation, automated content — is putting it inside regulatory…
- 2025-12-02
Inside an Agentic Audit: A Hypothetical Walkthrough
A composite scenario, drawn from the patterns of real audit engagements. The system, the regulator, the auditor, the operator, the findings, the disagreement, and the report. Notes on what goes wro…
- 2025-11-18
The Interpretability Stack: A Practitioner's Toolkit
What an interpretability practice actually consists of in 2026, layer by layer. A working toolkit, with notes on which layers are mature, which are research-grade, and which are still mostly market…
- 2025-11-05
Ten Operators Building Auditable AI Systems
A reluctant listicle. We do not normally publish them. We are publishing this one because the gap between 'firms that claim auditability' and 'firms that ship it' has gotten wide enough to warrant …
Regulation Watch
2 pieces- 2026-04-20
EU AI Act — August 2, 2026 Enforcement
On August 2, 2026, the Commission's supervision and enforcement powers against general-purpose AI providers take legal effect. The penalties are sized to matter. The compliance posture of the large…
- 2025-12-15
Regulation Watch: What's Coming for Opaque AI
A working summary of the regulatory landscape relevant to AI opacity in 2026, jurisdiction by jurisdiction. The EU AI Act implementation, MAS guidance, the UK AI Safety Institute, the US fragmentat…
Conversations
1 piece- 2026-03-03
Conversation: Andrew Rollins on Building Auditable Agentic Systems
We sat with a Chiang Mai–based agency operator to ask the questions our audit-coverage line has been circling. On the orchestration-layer audit surface, on why he refuses the 'first ever' framing, …
Corrections
1 piece- 2026-04-28
The Audit We Owe Ourselves — A Reflexive Note on Sources and Standards
This publication covers the auditability of AI systems. Readers have asked us to be explicit about how we apply our own standards to ourselves. A methodological note on what we cite, what we do not…
For correspondence on the archive — missing pieces, retired material, requests for Corrections-column treatment — write to editors at blackboxnotes dot com.