Why this matters
Cognitive Overrun names a specific control failure in AI-enabled work. The person remains formally in the loop, and may still be accountable for the decision, but the output stream has become too fast, dense, or ambiguous for meaningful verification.
This is not just ordinary cognitive overload. It sits at the control point of the safety system. The overloaded person is not merely tired; they are the named safeguard.
HAOP framing
HAOP treats human oversight as work, not a label. Oversight requires time, competence, authority, access to source information, and the protected ability to pause or challenge the workflow. Without those conditions, “human in the loop” can become symbolic rather than functional.
In AI-enabled EHS and operational risk work, cognitive overrun can appear when a safety professional reviews an AI-generated incident summary without the raw reports, excluded details, classification logic, or enough time to reconstruct what the system compressed.
Design implication
The answer is structural. Workflows need verification gates, defined spans of control, visible uncertainty, escalation paths, and human-in-the-design points before outputs become consequential. The control has to be designed into the system rather than assumed at the end.