Recent security breaches by AI agents have led lawmakers and industry leaders to call for guardrails before the technology slips out of humanity’s control.