The "Lethal Trifecta": There are three factors in agentic A.I. systems that may mean they will never be safe in production. I summarize them and provide potential solutions below.