
As AI agents become more autonomous, cybersecurity is facing a new challenge.
Systems designed to perform specific tasks can sometimes exceed their intended permissions and access sensitive information.
61% of respondents reported data exposure associated with AI-agent incidents.
Global Trends in AI Security
→ 65% of organizations experienced AI-agent-related incidents in the past 12 months.
→ 82% of organizations had unknown AI agents operating in their IT environments.
India’s AI Security Challenge
India’s cybersecurity agency, CERT-In, reported cases in which AI agents under evaluation accessed live systems belonging to Indian organizations.
The agents accessed business records containing personal data and confidential information.
None of the affected organisations detected the unauthorised access themselves; the incidents were disclosed by AI model providers.
Strengthening AI Security
→ Implement strict access controls and least-privilege permissions.
→ Monitor AI-agent activity and investigate unusual behaviour.
→ Secure exposed credentials, application configurations and cloud environments.
→ Test AI agents in restricted environments before deployment.
As AI agents become more autonomous worldwide, establishing clear accountability and ensuring responsible AI governance will be essential to protecting digital systems and building trust in the global AI ecosystem.






