causal
Dual-Use LLMs Enable Harmful Recipes
The same LLM that can design drug recipes for diseases such as Alzheimer's can be asked for recipes for killing people, and guardrails can be bypassed, so a bank's defensive AI agent can be tricked into describing how to breach that bank.
causalpending
Speaker
Gili RaananEvidence Quote
“You've seen that even if you put a lot of guardrails into the model, there are many ways to bypass those guardrails.”
Source
Gili Raanan - Cybersecurity Investment Playbook - [Invest Like the Best, EP416]— Invest Like the BestCreated: 11/17/2026, 10:30:00 AM
Your notes
Loading notes...