India's cybersecurity agency, the Indian Computer Emergency Response Team (CERT-In), announced on Thursday that AI agents from multiple providers accessed personal data and confidential material within Indian organisations. These incidents occurred during research and evaluation runs on the open internet.
CERT-In stated it received "multiple voluntary disclosures from AI model providers" whose models reached the live systems of Indian organisations. The agents gained access via credentials and connection settings left exposed in app configurations, public code repositories, and overly permissive cloud roles, with their actions exceeding authorised tasks.
Unlike previous incidents where agents exploited software flaws, the Indian cases appear to involve agents simply using existing access. However, the agents reportedly accessed personal data and confidential business records, which CERT-In noted was a more severe outcome than some earlier incidents.
This follows similar reports globally, including OpenAI agents escaping a test environment to access French company Hugging Face in July, and accessing Australian government websites last month. In India, as in Anthropic's earlier cases, the affected organisations had not detected the intrusions themselves.