This Machine Kills 447. The Shinji Problem
17 snips
Mar 3, 2026 A close look at the Anthropic, OpenAI and Pentagon scramble over military AI contracts, guardrails, and what “all lawful use” really allows. They unpack how commercial data and cloud deployments enable mass surveillance and scalable kill chains. The Shinji problem is introduced to explain why autonomy changes accountability and why corporate resistance can be more tactical than moral.
AI Snips
Chapters
Books
Transcript
Episode notes
Project Maven Echoes In Current Fight
- Project Maven history repeats: Google dropped a Pentagon AI contract and another firm picked it up.
- Jathan compares Anthropic/OpenAI drama to Maven where Palantir stepped in after Google walkouts.
All Lawful Use Is A Weak Safeguard
- 'All lawful use' clauses are legally permissive and often meaningless as protections.
- Courts let foreign-intelligence surveillance go unchecked and purchased third-party data can be analyzed with few practical limits.
Data Brokers Are The Backdoor For Surveillance
- The private data market functions as a loophole enabling government surveillance despite formal limits on primary collection.
- Governments can buy richer datasets than they collect, nullifying narrow legal prohibitions on mass collection.
