The AI agent once called ClawdBot is enchanting tech elites, but its security vulnerabilities highlight systemic problems with AI.
https://www.404media.co/silicon-valleys-favorite-new-ai-agent-has-serious-security-flaws/
@404mediaco I’m not sure it can be solved. How can you let a word interpreter run automatically without it just doing whatever. It doesn’t know which words are yours. It’s not just a problem for @steipete and perhaps people should be quiet about processing email with bots especially if you let them do things like purchases for you. I think prompt injection is not going away so just need to not be a target.