8 stories, newest first · one card per story
Non-privileged malicious native apps misuse the NativeMessaging API to intercept sensitive data or launch confused deputy attacks against browser extensions with over 532 million cumulative downloads.
The post treats injected content as a command channel and says evidence must come from logs the agent cannot change.
CVE-2026-103922 lets attacker-controlled content run with full origin trust inside Capacitor Android and iOS apps.
GitHub Security Lab’s AI taskflows uncovered 24 Android bugs, including OsmAnd location tracking.
A Kubernetes StatefulSet flaw lets namespace-scoped writers create a pod in another namespace.
A design flaw in Google's Kubernetes Config Connector allows attackers with namespace access to take over an entire GCP organization using a single YAML file.
Verifiable Action Card architecture blocks indirect prompt injection in agentic browsers, cutting attack success from 68-100% to 0%.
Noma Labs researchers describe 'workflow identity hijacking,' letting unauthenticated users trigger privileged AI workflows that execute actions with high-privilege service accounts.
Today 1.6K readers · 2.5K views · 778 reads · all time 189K views