According to the study, 54% of enterprises have already experienced incidents related to AI agents, and most of them still allow agents to share credentials. This creates significant risks for data and system security.
The study covered 107 enterprises and revealed that control measures necessary to restrict AI agent access are lagging behind their adoption. Only about one-third of enterprises assign a separate identity to each agent, and merely 30% isolate the highest-risk agents.
editorial commentary
Why it matters
The rise in incidents involving AI agents points to the necessity of improving security measures. The next signal could be an increase in the number of incidents or the implementation of stricter access management policies. However, uncertainty remains regarding how effective these measures will prove to be.