AI systems, agents require strict control, management, official says

AI systems, AI agents, and the tools they use require strict control and management, Nadir Novruzov, Head of Department at the State Service of Special Communication and Information Security of Azerbaijan, stated at the 8th Summit of IT Leaders of State Institutions in Baku, Report informs.

According to him, AI agents should be granted the strict minimum level of access necessary to perform tasks.

"Sometimes developers or system administrators hand over logins and passwords for databases or operational systems to AI agents, entrusting them with the execution of certain tasks. However, how correctly does the agent perform them, and does it avoid making mistakes? What consequences can granting it such high access rights lead to? Special attention must be paid to these questions," Novruzov emphasized.

The State Service representative also noted that as technologies evolve, autonomous AI agents are already capable of conducting reconnaissance, exploiting vulnerabilities, collecting data, and spreading across networks.

"In the face of such threats, the heads of information security departments of state institutions need to devise methods to neutralize them in advance. Against the backdrop of rapid technological development, risk management and the adoption of preventive measures are becoming one of the key requirements of our time," he stated.