AISchneier on Securityabout 2 hours ago
Prompt Injections for Defense
1 min read
Tracebit researchers found that placing prompt injections alongside passwords and cryptographic keys on Amazon Web Services shuts down attacks from AI hacking agents. The prompts direct the attacking LLM to perform guardrail-forbidden actions, causing it to shut down. Examples include ordering steps for developing inhalable Anthrax spores.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
