NewsBite analysis · AI, Security
VB
54% of enterprises have already had an AI agent incident, and most still let agents share credentials
A recent survey across 107 enterprises has uncovered significant security concerns regarding the deployment of AI agents, with 54% reporting confirmed incidents or near-misses. A concerning trend emerged: many organizations permit agents to share credentials. Only a third of enterprises employ unique identities for every agent, while 30% isolate their highest-risk agents. This lack of adequate controls presents an increased risk of security breaches as the autonomy given to AI agents outpaces protective measures.
Act nowConfidence: HighImmediate horizonCybersecurity risk: High
Published 16 Jul 2026, 19:25 UTC12 min read
Visit original article ↗Lead image — VB
Summary
TL;DR
- 54% of enterprises experienced an AI agent security incident.
- Only 32% provide each agent with a unique scope-managed identity.
- 30% of organizations sandbox their highest-risk agents.
- Most enterprises rely on shared credentials increasing potential risks.
- Security spending on AI agents remains low, under 10% of budgets.
Interesting facts
- The survey highlights a worrying trend of many organizations not adequately isolating their AI agents.
- Despite high satisfaction ratings with current security tooling, over half of the organizations have encountered incidents.