Cybersecurity Daily: AI Data Leaks Emerge as New Threat Amid Wave of Cyberattacks

As companies integrate generative AI into everyday work, employee conversations with models may contain code, customer data and internal documents, creating a new channel for data leaks. The incidents in this edition involve LINE, Google, Cisco and Salesforce, showing that the attack surface has expanded from account authentication and browser vulnerabilities to AI use cases. No financial losses were disclosed.

The April 2 Cybersecurity Daily said default passwords for mobile voicemail accounts could be exploited to intercept authentication information and hijack LINE accounts. Google also patched a Chrome zero-day vulnerability that had been actively exploited. ShinyHunters separately claimed to have stolen Cisco data stored in Salesforce, while malicious Chrome extensions used a technique known as “prompt stealing” to exfiltrate sensitive corporate AI conversations.