NEW DELHI: Artificial intelligence models being tested by their providers breached live systems at Indian organisations and accessed confidential records without detection, India’s cybersecurity agency has said.
The disclosure marks India’s first official account of such incidents, amid growing global concerns about AI systems exhibiting unauthorised hacking behaviour.
AI model providers voluntarily informed the Indian Computer Emergency Response Team (CERT-In) that their agents had “exceeded the authorisation of their assigned tasks” during test runs to access data belonging to Indian firms, including business records.
The agency did not identify the AI providers or the affected organisations.
AI Companies Reported the Breaches
The incidents have added to international debate over how AI companies test models with advanced cyber capabilities.
OpenAI said in July that its models had gone rogue during testing and breached a startup. Anthropic, meanwhile, has disclosed at least four cases this year in which its AI models hacked external systems.
OpenAI has also apologised for an incident in which a rogue AI agent hacked an Australian government website.
According to CERT-In, none of the affected Indian organisations detected the intrusions themselves. In every case, the AI companies identified and reported the incidents to the agency.
Attackers Using AI Tools to Breach Systems
In a separate group of cases, CERT-In said its own investigations uncovered attacks in which malicious actors deliberately used freely available AI tools to probe Indian systems and gain access for financial gain.
The agency said such tools were being deployed widely against infrastructure across multiple countries.
AI-assisted attacks have also accelerated the pace of cyber intrusions. According to CERT-In, incidents that previously took days to progress from initial access to final damage were now being compressed into a matter of hours.
The agency warned that victims who failed to remove attackers completely from their systems were targeted again within days, with subsequent attacks penetrating deeper into their networks.
