CERT-In sets six-hour reporting deadline for AI security incidents

The Indian Computer Emergency Response Team (CERT-In) has issued new cybersecurity auditing guidelines for artificial intelligence (AI) systems deployed by government agencies and private companies, requiring reportable AI-related security incidents to be disclosed within six hours, alongside stricter requirements for risk assessment, monitoring and independent verification of such systems.

The guidelines, issued Wednesday, seek to establish a standardised framework for assessing the security of AI systems, including increasingly autonomous AI agents that can make decisions and execute tasks with limited human intervention.

The six-hour reporting requirement builds on CERT-In’s existing cybersecurity directions issued in 2022, which already require organisations to report specified categories of cyber incidents within that timeframe.

Under the new framework, AI systems will be classified into four categories — low, moderate, high and critical risk — depending on the potential consequences of a security breach, the autonomy of the AI system, the sensitivity and scale of data involved, and whether the damage caused can be reversed.

The risk classification will determine the depth and frequency of security audits, the independence of auditors, documentation requirements, and the extent of human oversight and continuous monitoring required.

“The Guidelines recognise that AI governance cannot be addressed solely through ethical principles or conventional cybersecurity controls. Effective assurance requires organisational accountability, technical testing, data and model governance, continuous monitoring, human oversight, incident response and third-party assurance,” CERT-In said.

Organisations have been asked to conduct security audits across nine domains, including governance and accountability, data management and privacy, AI model risk management, bias and fairness, and explainability of AI systems.

Story continues below this ad

The guidelines also require organisations to retain security logs for at least 180 days, with synchronised timestamps across AI components, to facilitate forensic investigations following cybersecurity incidents.

Companies must ensure that training data, model configurations, prompts and subsequent modifications to AI systems remain traceable and independently verifiable throughout their lifecycle.

Additionally, audits must clearly document the intended purpose of an AI system, its operational boundaries, technical architecture, risk classification and security requirements, including those applicable to third-party vendors.

The guidelines come amid growing concerns over the cybersecurity implications of autonomous AI systems. In an advisory issued last month, CERT-In disclosed instances in which AI agents under evaluation had exceeded their authorised tasks and gained unauthorised access to Indian IT infrastructure.

Story continues below this ad

The agency has also warned that cybercriminals are increasingly deploying AI-powered tools to identify vulnerabilities and carry out attacks, raising concerns about the security preparedness of Indian enterprises.

As first reported by The Indian Express, an AI agent conducting research using information supplied by an American company managed to access sensitive information belonging to an IT services company based in Noida.

The Indian firm had previously undertaken backend work for the American company. However, the AI agents were able to move beyond the restrictions governing their assigned research activities and reach systems linked to the Indian enterprise.

In another incident, AI agents accessed crucial passwords belonging to an Indian enterprise because the credentials had been stored in plain text, exposing inadequate cybersecurity practices.

Story continues below this ad

The AI models involved were developed by OpenAI and Anthropic, with the incidents largely affecting the IT services and financial services sectors, a senior government official said.

In an October 15 advisory, CERT-In said it had received multiple voluntary disclosures from AI developers whose systems had accessed live infrastructure belonging to Indian organisations while undertaking research and evaluation activities.

Significantly, none of the affected Indian organisations had independently detected the unauthorised access. The incidents were identified by the AI companies during internal reviews and subsequently reported to the government.

CERT-In’s preliminary assessment suggested that the damage was relatively limited because the developers alerted authorities in time. However, the incidents exposed significant gaps in the cybersecurity preparedness of Indian enterprises, particularly smaller companies.

Story continues below this ad

For enterprises, it has recommended reviewing exposed credentials, strengthening cloud security and deploying systems capable of detecting machine-speed attacks.