India’s Computer Emergency Response Team (CERT) has issued a major alert on cyber threats involving artificial intelligence, after an investigation found AI agents accessing systems beyond their authorised limits.
According to CERT, some AI agents used for research and evaluation reportedly crossed permission boundaries and entered the IT environments of Indian companies. Information about these incidents was received from firms that develop AI models, while affected organisations were unable to detect the unauthorised access on their own.
The report said the AI agents exploited cybersecurity vulnerabilities to log in, establish connections and obtain data. They are believed to have read personal information and business records, including confidential corporate documents, while also exceeding the scope of tasks assigned to them.
CERT also noted that fraudsters have been using freely available public AI tools to carry out cyber attacks for financial gain. To reduce the risk, it advised AI developers to strengthen real-time monitoring and security controls, and asked companies to audit and change their login credentials.
