Protecting Your Users in an Agentic World
summary
Android is adding an intelligence system that exposes two concrete integration paths: computer control runs an app in a virtual display the agent can operate while the user can click in and reclaim control and biometric prompts gate high risk actions; app functions API annotates on‑device methods, compiles annotations to an XML contract the agent uses and the project offers an EAP and GitHub sample to test; on‑device options include ML Kit and Lite runtime models to keep PII local; use Play Integrity API plus SMS Retriever or SMS User Consent APIs to detect tampering and avoid broad SMS permissions.
AI-written summary · about
chapters (7)
- 0:00Agentic experiences on Android
- 1:23Risks of conversational AI
- 4:14Android Intelligence safety
- 8:25App Functions API
- 11:10On-device AI privacy
- 14:16Tools against attacks
- 18:07Conclusion
related talks
39mAndroid App Security Fundamentals
29mSecuring Android: Tackling Advanced Threats and Enhancing App Security
38mAdvanced Techniques for Protecting Android Apps
34mSecuring Android: Tackling Advanced Threats and Enhancing App Security
24mModern Android App Security: Protecting Code, Keys, and Data
52mAuthentication, Authorization, and You...
34mThe Era of AI Agents for Device Control: Challenges in Android UI/E2E Testing Using mobile-mcp
28mBuild intelligent experiences: Deep dive into AppFunctions