87% of Enterprises Say Their AI Controls Need to Work Better
What Happened — A Vanson Bourne survey of 3,600 decision‑makers at companies with >$500 M revenue across North America, EMEA, APJ and Japan reveals a widening governance gap in AI use. More than half report that AI systems—chatbots, predictive analytics, RPA, etc.—operate autonomously without routine human review, and 87 % say they need a better way to approve, monitor and control AI. Among fast‑adopting firms, 46 % have encountered security, privacy or compliance problems, and 25 % have lost visibility or control over AI agents’ permissions. Respondents also flag unverified AI output, unapproved tools and unclear decision ownership as common issues.
Why It Matters for Trust & Control Assurance — The findings expose a systemic shortfall in AI governance, directly testing the control objective of AI model risk management and oversight that underpins many frameworks (e.g., NIST AI RMF).
Who Is Affected — Financial services, healthcare, manufacturing, retail and other large enterprises that have deployed or are scaling AI.
Recommended Actions
- Define a formal AI model approval workflow with documented policies and evidence‑capture steps.
- Deploy continuous monitoring of AI model inputs, outputs and permission changes to maintain audit‑ready logs.
- Conduct periodic AI risk assessments aligned with the NIST AI RMF governance tier.
Technical Notes – The survey covers five AI categories (conversational agents, generative tools, predictive analytics, RPA, NLP). The primary risk vectors are unsupervised autonomous operation, lack of permission visibility, and unverified output usage. Source: https://www.helpnetsecurity.com/2026/11/06/autonomous-ai-oversight-governance/