Banks Deploy Read-Only AI Cybersecurity Assistants With Human Oversight

Financial institutions’ security operations centers are exploring AI to speed up alert triage, data aggregation and incident analysis. But mistakes in banks’ production environments, which involve customer data and core systems, can be costly. Exeter Finance and Tyson Foods therefore advocate initially treating AI as a read-only assistant, with employees retaining decision-making and execution authority.

At the 2026 RSAC Conference, cybersecurity engineers from the two organizations said AI should first be used to reduce workflow friction and should not act autonomously without human oversight. Otherwise, prompt injection could induce systems to damage production environments. They disclosed no investment figures for the deployments, emphasizing instead guardrails, continuous adversarial testing and human-in-the-loop mechanisms.