Cybersecurity pros warn that action-based AI agents bring more risk than simple chatbots, and users don't always control what these systems do.
Wendi Whitmore of Palo Alto Networks says hackers are already trying to exploit these vulnerabilities with hidden commands.
Even OpenClaw's creator, Peter Steinberger, admits there are risks: It's crucial that users understand what these systems can and cannot do, so everyone can better protect their data.