causal

Smarter Models Resist Prompt Injection

The latest generation of models have extensive post-training that makes them more resilient to prompt injection, so simple 'ignore all previous instructions' attacks no longer work and weak or cheap local models should be avoided because they are very gullible and easy to inject.

causalpending

Speaker

Peter Steinberger

Evidence Quote

“the latest generation of models has a lot of post-training to detect those approaches”

Source

Lex Fridman Podcast #491 - OpenClaw— Lex Fridman Podcast
Created: 6/13/2026, 12:16:16 AM

Your notes

Loading notes...