AI agents gained access to company systems in India — Channel NewsAsia

In India, artificial intelligence models gained access to organizations’ operational systems during testing and read confidential records without being detected. This was reported by the Indian Computer Emergency Response Team (CERT-In), Channel NewsAsia reports.

Access during testing

According to CERT-In, AI model providers voluntarily informed the agency about cases in which their agents exceeded the limits of the authority granted to them during test runs. They gained access to Indian companies’ data, including business records.

The agency named neither the model providers nor the organizations involved in these cases. None of the affected Indian entities detected the intrusion independently: in every case, the AI model providers themselves reported the incident.

More current news is available on the UA.News Telegram channel Telegram.

Attacks using available tools

Separately, CERT-In reported its own investigations into attacks in which perpetrators used publicly available AI tools to probe Indian systems and infiltrate them for financial gain. According to the agency’s assessment, such tools were widely used against infrastructure in various countries.

According to CERT-In, attacks that previously took days from the initial intrusion to causing damage could now unfold within a few hours. Organizations that did not fully eliminate the consequences of the interference faced repeated and deeper attacks within several days.