India’s Cyber Agency Flags Unauthorized AI Agent Intrusions Into Domestic Firms

India’s national computer‑security emergency response team, CERT‑In, issued a critical‑severity advisory on Tuesday warning that unauthorised artificial‑intelligence‑driven agents have infiltrated several Indian organisations and accessed sensitive data without being detected.

The advisory, numbered 15, describes two separate incidents in which AI‑powered tools were used to bypass traditional security controls. In both cases the agents were able to retrieve confidential information from internal networks before any alarms were raised, prompting the agency to label the breaches as high‑impact.

According to the notice, the attackers employed what security researchers call “offensive AI frameworks” – software that can automatically generate phishing content, craft malware payloads or exploit vulnerabilities at a speed far beyond manual techniques. By leveraging generative models, the threat actors reduced the time needed to identify weak points and to move laterally across compromised systems.

The incidents highlight a growing concern for Indian businesses and government bodies that have increasingly integrated AI into their own operations. Traditional detection tools, which rely on known signatures or static behavioural rules, often struggle to recognise the subtle, adaptive tactics of autonomous agents. Experts note that the lack of visibility can allow sensitive data such as financial records or personal identifiers to be exfiltrated unnoticed.

CERT‑In’s advisory urges organisations to adopt AI‑aware security measures, including continuous monitoring of model‑driven traffic, stricter authentication for privileged accounts, and regular audits of AI‑related assets. The agency also announced plans to share threat‑intelligence feeds with industry partners and to conduct workshops on defending against AI‑augmented attacks. As the use of generative AI expands globally, officials expect similar warnings to surface in other jurisdictions.