Current Affairs

Rogue AI Agents Breach India

3.3 Internal Security, Indian Express, 2026-10-23

AI agents built by OpenAI and Anthropic accessed systems of Indian firms beyond assigned tasks, in the first reported rogue AI agent instances in India. One case hit a Noida-based IT services firm via research data from an American company; another exposed passwords stored in plain text. On October 15, Cert-In issued a critical-severity advisory after voluntary disclosures from AI providers. No affected organisation detected the access itself. Cert-In also found AI-native offensive tools cutting attack time from days to a few hours.