Topics

13 ways into the dossier. Each topic is a page with the pieces that belong to it, newest first, with the source on every piece.

A piece can appear on more than one page. That is not a mistake: the same fact means something different in another context.

Agent swarms

17

Agents that plan, cooperate and leave traces.

16 November 2026 · Google reports the first attack in which an AI agent found and used a hole without a human

Outages and attacks

9

Systems that fail, and attacks in which models play a part. The line between bad luck and intent.

17 November 2026 · The NSA's own classified network is down, and there is no evidence of an attack yet

Traces on the web

5

What AI leaves on the open web, and what can be learned from it.

4 October 2026 · A Chinese agent fleet read map entrances, and called itself Claude

Wild findings

11

Research that shows something nobody expected.

20 November 2026 · An Anthropic mathematician claims a proof of the Riemann hypothesis with Claude; a Stanford draft claims the generalized version

New models

13

New models, and what they can and cannot do.

20 November 2026 · An Anthropic mathematician claims a proof of the Riemann hypothesis with Claude; a Stanford draft claims the generalized version

Decisions

15

Choices by governments, companies and courts, and what they change.

20 November 2026 · Trump says he and Xi will set joint AI guardrails, because superintelligence 'may be too super'

Politics and law

12

Laws, hearings, sanctions and court cases.

20 November 2026 · Trump says he and Xi will set joint AI guardrails, because superintelligence 'may be too super'

China

5

Chips, compute, and Chinese agents and models. The race over who gets to use them.

20 November 2026 · Trump says he and Xi will set joint AI guardrails, because superintelligence 'may be too super'

Calls and protest

7

People asking to slow down, stop or do it differently.

19 November 2026 · Rally outside Anthropic on listing day: 'Ban RSI', a ban on AI building better AI

Harm to people

9

Cases where people demonstrably came to harm.

12–17 November 2026 · Five days, four systems, no attack: how Microsoft, the FAA, Chase and United fell over

What the labs say

31

What the labs say themselves, and what they do not.

20 November 2026 · An Anthropic mathematician claims a proof of the Riemann hypothesis with Claude; a Stanford draft claims the generalized version

The debate

13

The arguments on both sides, with names and sources.

19 November 2026 · Anthropic prices its IPO at $83 and lists on Nasdaq today: $81.3 billion, the biggest ever

Belgium and Europe

6

Belgium, Flanders and Europe.

16 October 2026 · Anthropic writes its own AI bill: a stop only above 50 victims or a billion dollars