Open-source Android AI agents can turn hidden screen text into host commands, and all five tested frameworks failed at least six lab attacks.

Reported by The Hacker News.

Read the full story →