Survey Finds Most Firms Hit by AI Agent Incidents
Gartner's June and July 2026 survey of 298 infrastructure and operations leaders in the United States and United Kingdom found widespread agent-related problems.
Updated on Nov. 18, 2026 in Artificial Intelligence

Live Poll
Do you trust companies to keep AI tools under proper control?
Gartner surveyed 298 infrastructure and operations leaders in the United States and United Kingdom in June and July 2026. The survey found 95% of organizations had at least one incident involving an AI agent, a software tool that takes actions to complete a task, in the previous 12 months.
Why it matters
Only 15% of respondents had an automatic way to stop an agent when its errors got out of control, and 57% described their AI rules and controls as weak. Gartner projects that by 2028, 40% of IT teams expanding AI use without strong controls will face service disruption.
Gartner's survey of 298 leaders in the United States and United Kingdom found 15% had an automatic kill switch for runaway agents, while 55% called their IT risk management tools completely outdated. The survey does not detail the technical causes of the reported agent errors.
The players
Gartner
Gartner is a global research and advisory firm that publishes technology market research and conducts annual IT conferences.
The details
Gartner analysts are presenting the findings at the Gartner IT Infrastructure, Operations & Cloud Strategies Conference in London. The firm recommends keeping an inventory of AI agents, tracking each step they take, and requiring human approval for actions that carry serious risks.
Timeline
June and July 2026: Gartner conducts its survey of 298 infrastructure and operations leaders.
Previous 12 months: 95% of surveyed organizations report at least one AI agent incident.
2028: Gartner projects 40% of IT teams expanding AI use without strong controls will face service disruption.
The Tech Race
The survey points to a gap between rapid enterprise deployment of AI agents and the governance tools meant to control them. Gartner's warning about 2028 disruption frames governance as a competitive and operational priority rather than a later concern.
Teams using AI agents without automatic shutoffs or approval rules may face service outages that affect customers and internal operations. Organizations can reduce exposure by inventorying their agents and requiring human sign-off for high-risk actions.
The takeaway
Most organizations are already experiencing AI agent incidents, so governance controls such as kill switches and agent inventories are becoming operational necessities. Building them now can prevent costly disruptions later.
Further reading
For more on how companies are managing these systems, see our Artificial Intelligence coverage.
Source note: This article includes information reported by Back End News.
Live Poll
Do you trust companies to keep AI tools under proper control?







