Ideas & arguments / October 2026
Private AI, independent audits, and peaceful coexistence
We research how Fully Homomorphic Encryption could keep AI conversations private and protect records for independent model audits. These essays explain the technical proposals and the questions of power behind them.
Our central question is who controls powerful intelligence: who may think privately, collaborate, inspect evidence, and help set the rules. The arguments and proposals are ours; the cited studies are the work of their authors.
Peaceful Coexistence With AI
How can humans and increasingly capable AI cooperate safely? Fair procedures, private thought, and independent evidence give disagreements a peaceful path to resolution.
Fully Homomorphic Encryption and the Right to Think
Who can read an unpublished idea—and who gains power when multiple organizations can use AI privately? Our case for privacy, and the risks it creates.
Move the Audit Outside the Model
A model behaves differently when it encounters an audit cue. What could separately recorded calculations tell an investigator?
Can Intelligence Outgrow Cryptography?
Could lattice-based cryptography protect privacy against quantum computers and future superintelligence? We explore the link to learning and quantum cognition.