Security in AI operations
Starting point
AI agents read company data and touch systems; prompt injection, data exfiltration and missing logs go untreated.
Consequence if unaddressed
Consequence: an incident via the agent is not traceably covered — liability and notification-duty risk.