Learning Center
What happens when AI acts?
Films, the Consequence Library, the law as it changes, and the proof. Start anywhere.
The library
Every shelf.
AI agent security
For CISOs and security teams: what agents have done, what the numbers say, and the words in plain English.
AI agents and insurance
For carriers, brokers and underwriters: the stop, the record, the rules and the exclusions.
- The stop is now insured. Prove who pressed it.
- AI and insurance in numbers
- An AI agent loss: what the claim will ask
- The AI agent underwriting question set
- The AI rules for insurers
- AI insurance in 2026: cover, exclusions and the rules for insurers
- The ISO generative AI exclusion, CG 40 47
- A driving record, for AI
Research library
Reports compiled from public sources: incidents, life sciences and GxP, sectors, the law, and human oversight.
Human authority in the loop
What it means, and how it differs from a human in the loop.
The Cleanroom Standard for AI
Written by Louisa Johnson Bullock. What a governed AI deployment looks like, in three classes an outsider can verify.
Research notes
What the published record shows, with every source linked.
- Why AI goes rogue, and why drift is rising
- The human cost of approving AI agent actions
- AI insurance in 2026: cover, exclusions and the rules for insurers
- AI agents in pharma production: what has been announced
- Colorado’s draft rules on human review of automated decisions
- AI law, June to September 2026: what changed
- Every report in the Research library
Films
Every idea on this site, in about a minute.
- 01 · Refund every customer
- 02 · An AI drifts on its own
- 03 · Kevin drifts
- 04 · One AI agent asks another for your photos
- 05 · Two companies, one batch
- 06 · No need to slow down
- 07 · An AI goes off procedure
- 08 · Your AI, on their radar
- 09 · Ask the Transponder™ what your AI did
- 10 · One engine. Seven proofs
- 11 · AI agents, an insurance problem
- 12 · Kevin and Gary, the plumber
- 13 · Kevin and Gary: the school pickup
- 14 · Don’t be like Kevin and Gary
- All films, sorted by what they show
Kevin and Gary
The first home robots, 2031. They say yes to everything, and demonstrate the problem.
The Consequence Library
AI and automated systems that acted, and what followed. Sourced and graded.
- Knight Capital: 45 minutes, $440 million, no kill switch
- Replit agent deletes a production database in a code freeze
- PocketOS: a database and its backups, gone in nine seconds
- Moffatt v. Air Canada: a chatbot commits the company
- The first FDA warning letter on AI
- An agent deletes an inbox and ignores stop commands
- Robodebt: automated debt raising at population scale
The law as it changes
What the rules actually say, and what they ask a record to show.
- Who requires a record of your AI use
- The laws and standards
- Federal AI standing orders
- AI rules for lawyers, country by country
- Annex 11 and AI
- GDPR Article 22: meaningful human intervention
- Model risk and staff AI use
- Local Law 144 bias audits
- California SB 574: signed, Chapter 858 of 2026
- When the law changes
AI in court
The sanctions, the cases and the privilege split.
The Drift Desk
Drift, in plain words: what it is, why it happens, and the off switch.
Proof and the record
What a defensible record shows, and how anyone can check it.
Research
Published tests behind the words on this site.
