AI security notes 6/27 (opens in new tab)

Discussed on Substack

AI generated code security risks stem from basic LLM limitations; we should frame prompt injection models like other detection and response tools