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

Google Security Blog coverage highlights active and emerging security concerns including indirect prompt injection against AI agents and Workspace integrations, browser session-cookie theft and defenses using Device Bound Session Credentials, memory-safety risks in cellular modem firmware, post-quantum cryptography migration for Android and HTTPS, and ongoing vulnerability research through Google’s VRP. The material is primarily defensive and strategic, but prompt injection, credential theft, and remotely exploitable modem vulnerabilities represent significant threat areas.