Topic

LLM jailbreaks

2 mentions in 1 episode

The hosts discuss tests in which simple attacks elicited harmful responses from large language models.

Mentions

649: Garbage In, Garbage Out

  • 10:05

    four undisclosed LLMs tested were highly vulnerable to basic jailbreaks

  • 10:24

    these guardrails that they've put on these major LLMs are basically not working