AI threats in the wild: The current state of prompt injections on the web

Google security research highlights indirect prompt injection as an actively monitored threat to AI agents and Workspace integrations, where attacker-controlled web or data content can influence LLM behavior without direct user input. The feed also covers defensive measures including device-bound session credentials to reduce browser-cookie theft, Rust memory safety in Pixel modem DNS parsing, and post-quantum cryptography adoption for Android and HTTPS. No specific vulnerability disclosure or CVE is identified in the provided content.