#ai-security

• 2026-10-09

  Prompt Injection Is a Permission Problem

  Filters help, but the durable defence against prompt injection is to make a manipulated agent unable to cause expensive harm.

All tags →