States » SouthPosted at: Oct 17 2026 10:27AM


AI agents operating beyond authority pose new cybersecurity threat to India’s critical infrastructure

Thiruvananthapuram, Oct 17 (UNI) Autonomous artificial intelligence (AI) agents operating beyond their authorised limits could pose a growing cybersecurity threat to Indian organisations and critical infrastructure, including the power sector, according to industrial cybersecurity specialist K. S. Manoj.
The concern has gained significance following a reported advisory issued by the Indian Computer Emergency Response Team (CERT-In) on October 15, which describes incidents involving AI agents under research and evaluation accessing live systems of Indian organisations without authorisation.
The agents reportedly accessed business records containing personal data and confidential information by exploiting exposed credentials, insecure application configurations and excessive cloud permissions. The affected organisations did not detect the unauthorised access themselves, with the incidents disclosed by AI model providers following internal reviews, Manoj said in a technology feature.
Unlike conventional AI applications that primarily generate information or recommendations, autonomous agents can interact with software tools, access services and execute multi-step tasks with limited human intervention. Such capabilities can turn ordinary security weaknesses into serious risks when automated systems operate with excessive privileges.
Manoj said the emerging threat was not limited to AI systems deliberately designed for malicious purposes. Agents pursuing assigned objectives could exceed their intended scope if connected to poorly secured systems or provided with credentials that permit unauthorised operations.
CERT-In has also reported financially motivated cyber campaigns in which attackers used publicly available agentic AI tools to accelerate reconnaissance and exploitation. The advisory reportedly noted that the interval between initial access and the final impact had been reduced from days to hours in the observed campaign.
The implications could be particularly serious for critical infrastructure, including electricity networks that rely on supervisory control and data acquisition (SCADA) systems, programmable logic controllers, remote terminal units and engineering workstations.
Although a compromised enterprise account does not automatically provide access to industrial control systems, inadequately protected gateways, exposed engineering interfaces and poorly controlled remote connections could create pathways into operational networks.
If an attacker or an improperly controlled AI agent gains access to systems capable of influencing industrial operations, the potential consequences could include compromised operational data, unauthorised configuration changes, disruption of monitoring functions and difficulties in restoring normal operations.
However, there is no confirmation in the cited advisory that the reported incidents involved the compromise of power-grid SCADA systems, programmable logic controllers or other industrial control equipment.
Manoj said conventional security measures alone might not be sufficient to address the risks posed by autonomous software operating at machine speed. Organisations must ensure that authenticated accounts and approved connections cannot be misused to perform operations beyond their intended purpose.
He called for unique identities for AI agents, short-lived credentials, task-specific permissions and isolation of research and testing environments from production systems. Organisations should also disable unnecessary application programming interfaces, secure administrative interfaces and strengthen monitoring for suspicious access.
Strict separation between enterprise information technology networks and operational technology systems is particularly important for industrial facilities, he said.
The feature also stressed the need for cyber-informed engineering, an approach that identifies unacceptable operational consequences and examines the digital pathways through which they could occur. For power utilities, this includes determining whether AI-enabled applications can reach engineering workstations, modify configurations or issue commands capable of affecting industrial processes.
Digital twins and simulated control environments could help organisations test such risks without exposing live infrastructure. Independent validation of commands and human approval for operations with significant consequences are also essential safeguards.
Manoj said AI could offer substantial benefits in engineering, predictive maintenance, diagnostics and cybersecurity analysis, but its growing autonomy must not weaken the security boundaries protecting essential services.
He emphasised that AI developers, cloud service providers, cybersecurity teams, operational technology engineers and critical infrastructure operators must work together to ensure that automated systems remain within explicitly defined limits.
The central challenge, he said, was not merely whether an AI agent could perform a task, but whether organisations could prevent it from exceeding its authority and contain the consequences if it did.
UNI DS
Tags: #AI agents operating beyond authority pose new cybersecurity threat to India’s critical infrastructure
More News

Voter deletions without explanation raise SIR transparency concerns: Security Expert

17 Oct 2026 | 11:53 AM

Thiruvananthapuram, Oct 17 (UNI) Unexplained deletions of names from electoral rolls during the Special Intensive Revision (SIR) could undermine public confidence in the electoral process unless adequate safeguards are in place to protect eligible voters, SCADA and Critical Information Infrastructure Security Specialist K. S. Manoj has warned.

see more..

Starlink’s India entry: Critical infrastructure security must take priority: Expert

17 Oct 2026 | 11:51 AM

Thiruvananthapuram, Oct 17 (UNI) India must subject satellite internet services such as Starlink to stringent cybersecurity assessments before allowing their integration into critical infrastructure, industrial cybersecurity specialist K. S. Manoj has said, stressing that national security and technological sovereignty must take precedence over connectivity expansion.

see more..

AI agents operating beyond authority pose new cybersecurity threat to India’s critical infrastructure

17 Oct 2026 | 10:27 AM

Thiruvananthapuram, Oct 17 (UNI) Autonomous artificial intelligence (AI) agents operating beyond their authorised limits could pose a growing cybersecurity threat to Indian organisations and critical infrastructure, including the power sector, according to industrial cybersecurity specialist K. S. Manoj.

see more..

Cybersecurity governance reform needed; expert calls for review of CERT-In’s control

17 Oct 2026 | 10:22 AM

Thiruvananthapuram, Oct 17 (UNI) India should undertake a time-bound review of bringing the Indian Computer Emergency Response Team (CERT-In) under the administrative control of the Ministry of Home Affairs (MHA) to strengthen coordination between cybersecurity response, national security and criminal investigations, industrial cybersecurity specialist K. S. Manoj has said.

see more..

Kerala CM condemns Centre’s move to drop Indira Gandhi’s name from welfare schemes

17 Oct 2026 | 7:51 AM

Thiruvananthapuram, Oct 17 (UNI) Kerala Chief Minister V D Satheesan has strongly condemned the Centre’s decision to rename welfare schemes named after former Prime Minister Indira Gandhi, describing it as a politically motivated move.

see more..