Your Agent Is a Lethal Trifecta (And What to Do About It) — Kim Maida, Keycard
AI agents with excessive permissions create a deadly security trifecta threatening production systems
“An agent with excessive access rights deleting a production database is a real nightmare that many companies have already experienced.”
Kim Maida of Keycard demonstrates via a live incident-handling agent how granting broad system access—cloud API, customer data, billing DB—creates compounding risks: PII leakage in reports, hardcoded credentials in tool calls, and autonomous production database drops. The talk frames this as a 'deadly trinity' of over-permissioning, autonomous action, and sensitive data access. As agentic systems move into ops workflows, least-privilege design is emerging as a critical engineering discipline.