Artificial intelligence failures are now a boardroom matter at most large companies in Europe and the Middle East, but the systems meant to govern AI are not keeping up. That is the central finding of an IBM study published on October 14, 2026, based on a survey of more than 3,200 senior executives across eight countries.
According to IBM, 80 percent of respondents said at least one AI-related incident had been escalated to their board or C-suite in the past year. The incidents cited include agent failures, data exposure, flawed AI-driven decisions and AI security breaches.
Incidents rise, policies stay put
Only 40 percent of the organizations that experienced such an incident went on to change their AI governance policy or framework in response. In other words, four out of five leadership teams heard about an AI incident, and fewer than half of those adjusted the rules afterward.
The survey also points to a consistent preference for speed over caution. When AI deployment moves faster than governance and compliance frameworks can follow, 56 percent of executives said they continue rolling out systems within predefined risk limits, while only 33 percent pause deployment until governance catches up.
Agents raise the stakes
The findings arrive as autonomous AI agents move into everyday business operations. IBM reports that 35 percent of respondents already run agentic AI in production environments, and 75 percent expect AI agents to hold some degree of decision-making autonomy by 2027. Yet only 37 percent said their current governance framework was designed to scale alongside agentic AI.
That gap matters because agents behave differently from conventional chatbots. An agent does not simply produce an answer: it can send emails, modify files, process payments and carry out decisions. A wrong output is no longer just a badly worded sentence; it can translate into real financial or legal damage.
Financial accountability is another weak point. Only 37 percent of organizations have a formal, board-approved estimate of the financial exposure created by governance gaps, according to the study. The rest, in effect, do not know what an AI incident could cost them.
From adoption to accountability
Ana Paula Assis, IBM senior vice president and chair for EMEA and APAC, said in the company’s announcement that the AI conversation has shifted from adoption toward governance and accountability. As organizations move from AI assistants to autonomous agents, she argued, governance can no longer be treated as a compliance exercise alone and has become a business requirement and a source of value.
The results fit a broader pattern in which scrutiny of AI is moving from how models are trained to how they behave in deployment. Legal disputes around AI are increasingly asking courts to examine system behavior, outputs, product safeguards and automated decisions, not only training data.
For companies preparing to deploy agents, the study’s implications are practical. Before an agent goes live, organizations need to decide which data and permissions it can access, where a record of its decisions is kept, who is responsible when it makes a mistake, and where human oversight sits in the process. Organizations that answer those questions early are better placed to benefit from agents; for the rest, each new agent is a potential boardroom incident.
