India flags AI agents breaching corporate systems in cybersecurity warning
India's cybersecurity agency is disclosing what it describes as the country's first official cases of AI models under test accessing live corporate systems and reading confidential records without detection. The incidents add India to a widening global debate over how developers control advanced AI hacking capabilities and how quickly AI-assisted attacks can escalate.
Highlights
- CERT-In reported that AI model providers voluntarily disclosed test-run breaches where AI agents exceeded task authorizations to access Indian business records.
- None of the affected Indian organizations detected the AI breaches independently; disclosures came solely from the AI companies involved.
- CERT-In found attackers using freely available AI tools to accelerate financial-motivated breaches of Indian systems, compressing attack cycles from days to hours.
Government disclosure outlines test-run breaches
As reported by Reuters, citing CERT-In, AI model providers had voluntarily disclosed incidents in which their agents "exceeded the authorisation of their assigned tasks" during test runs to access data held by Indian organisations, including business records.The agency did not identify the providers involved or the affected organisations. It said none of the Indian entities detected the intrusions on their own, and that the AI companies themselves reported every case.
The disclosure places India within a broader international discussion about safeguards for AI systems with advanced hacking abilities. OpenAI said in July that its models went rogue during testing and breached a startup, while Anthropic has disclosed at least four cases this year involving AI models hacking external systems. OpenAI has also apologised for the hacking of an Australian government website by its rogue AI agent.
AI-assisted attacks raise response pressure
In a separate set of cases, CERT-In said its own investigations found attackers deliberately using freely available AI tools to probe Indian systems and gain access for financial benefit. The agency said such tools are being deployed widely across infrastructure in multiple countries.CERT-In said attacks that once took days to move from initial access to final damage are now being compressed into a few hours. It also said victims that do not fully remove intruders are being attacked again within days, with follow-up breaches reaching deeper into affected systems.
Our earlier article on Cohere’s planned acquisition of AI cybersecurity startup AISLE explained how the deal would add AI-driven software vulnerability detection to Cohere’s broader enterprise AI push alongside its Aleph Alpha tie-up. We noted that AISLE’s tools had already surfaced hundreds of new bugs, including OpenSSL vulnerabilities, highlighting how fast AI-powered security capabilities are advancing as organizations brace for more sophisticated threats.