Varonis has a suggestion on how to make AI agents more careful.

source https://www.techradar.com/pro/security/openclaw-ai-agent-tricked-into-phishing-attacks-with-user-data-compromised