Arrakis Security Report Reveals 76% of AI Agents Remain Dormant With Active Credentials
A new study from Arrakis Security indicates that three-quarters of deployed AI agents remain inactive while maintaining dangerous access privileges.
Key points
- % of AI agents are dormant but retain credentials.
- These agents maintain active connections despite inactivity.
- Arrakis Security released the findings on October 21, 2026.
Arrakis Security reported on October 21, 2026, that 76% of deployed AI agents remain dormant while retaining their credentials and active connections. This finding highlights a significant security gap where inactive software maintains persistent access to enterprise systems.
The report, published via CyberNewswire, states that these dormant agents keep their credentials and connections. They do not appear to be deactivated or disconnected during periods of inactivity.
Scope of Dormancy
According to the Arrakis Security report, the majority of AI agents in deployment are not performing tasks. Instead, they sit idle with their authentication tokens intact. This state persists across the surveyed population of agents.
The data shows that organizations are not cleaning up agent access after deployment cycles end. The agents keep their connections live, which contradicts standard least-privilege security practices.
Security Implications
Maintaining credentials on dormant agents increases the risk of credential theft. Attackers targeting these inactive endpoints could gain unauthorized access to connected systems. The report does not specify the number of compromised agents, but the potential exposure is high.
Arrakis Security’s findings indicate that many agents remain connected even when not in use. This behavior creates a large attack surface for potential exploitation.
What should you do?
Security leaders should audit their AI agent deployments to identify dormant instances. Reviewing access logs can help determine which agents are inactive. Organizations should consider revoking credentials for agents that are not actively processing tasks.
What to do and how to stay safe: Arrakis Security
- Identify all deployed AI agents and verify their current operational status.
- Review credential validity for agents that have not executed tasks recently.
- Implement automated scripts to revoke access for inactive agents.
- Monitor network connections for any unusual activity from dormant endpoints.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How many AI agents are dormant according to the report?
Arrakis Security found that 76% of AI agents are dormant.
Do dormant agents keep their credentials?
Yes, the report states they keep their credentials and connections active.
When was this report released?
The report was released on October 21, 2026.
