🚨 ServiceNow Warns: Uncontrolled AI Agents Threaten Companies! 🚨

[Süeda Asil]

Are AI Agents Running Out of Control?​

According to recent research by ServiceNow, the proliferation of uncontrolled artificial intelligence (AI) agents within companies is increasing risks as they transition from pilot phases to system integration. However, it is noted that some effective control mechanisms are already in place within the manufacturing sector.

Global Security Leaders Concerned​

According to a survey conducted by ServiceNow, a cloud-based AI governance software platform provider, with global security leaders, 90% of organizations worldwide experienced an AI agent-related security incident in the past year. The report, titled "Distributed but Undisciplined: The State of Agent Security," includes insights from over 3,700 senior security and technology leaders across 20 countries and eight sectors. Of these leaders, 575 are from the global manufacturing sector.

Deficiencies in Governance Frameworks​

The research reveals that less than half of organizations use governance frameworks that cover AI agent deployments, which increases the risks of uncontrolled AI agents. Most AI agents are managed with outdated frameworks and human-centric systems. Less than 50% of global security teams incorporate security elements into an agent's development.

According to ServiceNow, permissions and security measures are applied "as needed" rather than before deployment. This situation, combined with blind spots and scale, increases organizations' risk exposure. Furthermore, machine-speed risks outpace human-speed responses: While 76% of organizations state they maintain an up-to-date map of each agent's access, only 19% can see more than 90% of their agents in real time.

More Robust Security in the Manufacturing Sector​

Despite these challenges, the study found significant differences in the manufacturing sector. For example, 92% of manufacturing organizations have AI agents interacting with OT or physical environments, compared to 82% globally. In the manufacturing sector, over 30% detected previously unknown or unrecorded AI agents operating in production. Globally, this rate is 29%.

Despite increased exposure, only 63% of respondents in the manufacturing sector experienced an OT or physical security incident involving an AI agent, compared to 76% globally. ServiceNow suggests this may be due to the control mechanisms implemented by manufacturers. Specifically, 27% of organizations in the manufacturing sector can see more than 90% of their agents in real time, compared to 19% globally. Additionally, 39% of the manufacturing sector has an automated gate that prevents deployment until an agent passes security controls, compared to 32% globally.

5 Recommendations from ServiceNow​

ServiceNow offers five recommendations for organizations to strengthen their security:

• Create a continuous inventory and map agent access.
  
• Integrate security before the design phase.
  
• Consolidate visibility across agents, identities, and posture.
  
• Collect continuous evidence.
  
• Appoint a leader across the organization.