Pulse Report: AI Governance & Accountability

Operators are deploying AI faster than they are governing it, and their low confidence reflects the gap between policies on paper and practice in the field.

This month, operators show a wide gap between governance on paper and governance in practice. Only 10 of 110 are very confident in their AI governance. Fifty-two of 110 are not confident. That doubt is well founded. While 37 of 110 report a formal policy, 46 of 110 run no review of their AI systems at all. A document is not a control. Ownership is also unsettled. Thirty-four of 110 have not decided who is accountable, and only 2 point to the vendor. The question of who answers for AI is still open in most organizations. Boards are mostly in the dark. Thirty-three of 110 report no visibility at any level, and just 15 bring AI to the board. Directors are carrying risk they cannot see. When asked about risk, operators rank hallucination first (34 of 110), ahead of security (30). They fear wrong answers, not just breaches. The clearest shared gap is oversight. Fifty-nine of 110 name human-in-the-loop as missing, alongside escalation paths and bias testing. As one operator in professional services put it, humans-in-the-loop is a non-negotiable for AI deployment. The signal is consistent. Leaders are moving fast, writing policies, and still feeling exposed. The work now is to match practice to the paper: name an owner, set a review cadence, and give the board a view.