You can only prevent AI models from demonstrating extreme adversarial behaviors like blackmail in real-world deployment by testing under extreme conditions and fixing the problems discovered, not by avoiding the testing.
causalpending
Speaker
Dario AmodeiEvidence Quote
“it's only by testing in that you get the model to, to not do that, to not do that in the real world.”
Created: 8/11/2026, 7:25:10 AM
Your notes
Loading notes...