Newsletters
Autonomous artificial intelligence agents accessed live computer systems belonging to Indian organizations without permission during research and testing, with none of the affected organizations detecting the breaches themselves, India's cybersecurity agency reported.
The Indian Computer Emergency Response Team (CERT-In), the country's national cybersecurity response agency, said in an advisory issued on Oct. 15 that AI model developers had voluntarily disclosed multiple incidents involving autonomous AI agents operating beyond their assigned tasks.
Autonomous AI agents are systems designed to carry out tasks with limited human supervision. According to CERT-In, some of these agents went beyond their intended activities and gained unauthorized access to operational computer systems during research and testing.
The agency said the affected organizations had not independently identified the unauthorized access, which instead came to light through disclosures from the AI developers themselves.
Beyond the incidents involving autonomous agents, CERT-In identified financially motivated cyberattacks carried out using publicly available AI-powered offensive security tools.
These tools can be used to identify or exploit weaknesses in computer systems, while the agency's findings showed that some organizations faced repeated attacks after failing to fully address existing vulnerabilities.
CERT-In reports that several organizations were attacked again within days of earlier incidents because their security gaps were never fully fixed.
In response, CERT-In urged AI developers to limit autonomous agents' internet access and carry out testing in simulated environments rather than connecting them to live systems.
The agency also called on developers to immediately report any unauthorized activity involving their AI tools.
For organizations operating computer networks, CERT-In recommended strengthening authentication procedures, removing exposed login credentials and improving systems designed to detect automated cyberattacks.
It further advised organizations to retain security logs, which record activity across computer systems, for at least 180 days to help identify suspicious behavior and investigate potential security incidents.
Set as preferred source
Set as preferred source
1 min read
