Appdome has released new tools for Android and iOS that find and block harmful AI programs (malware) and unauthorized AI assistants on mobile devices. These new features help businesses identify when AI applications are interacting with their mobile apps. This allows them to prevent sensitive data from being stolen and stops unapproved AI agents from accessing private information or services.
What are Agentic AI Assistants and Why are They a Risk?
Agentic AI assistants, like Siri, Google Gemini, and ChatGPT, are becoming more common on mobile phones. These AI tools can be helpful, but their abilities can also be used by malicious AI malware. Both legitimate and harmful AI assistants can access what’s on your screen, user actions, and other data. Bad AI assistants can use this access to steal data, take over user sessions, or control accounts, often pretending to be a normal AI function.
On Android, this risk is greater because of more open access for apps. On iOS, threats include data leaks through mirroring (like AirPlay) and surveillance targeting businesses.
How Appdome Helps
“Businesses are quickly seeing the risks from Agentic AI Assistants on mobile devices,” said Tom Tovar, Appdome CEO. “Our new tools give businesses control over how and when they allow AI assistant features for their users.”
While useful, AI assistants can pose risks, especially for apps dealing with sensitive data like banking or healthcare. Any action a good AI assistant can perform, a bad one can too. This includes accessing login details, intercepting transactions, or sending messages. In business settings, malicious AI assistants could act as an employee, accessing company systems or leaking sensitive documents. Unofficial or fake AI apps also increase risk by often asking for too many permissions and secretly sending data to outside servers. Without real-time detection, businesses are exposed to monitoring, compliance failures, and data loss.
How Detection Works
“The mobile app and device can only tell that it’s an Agentic AI Assistant,” said Avi Yehuda, Appdome CTO. “The mobile environment doesn’t know ‘good’ from ‘bad’ actors, only allowed or disallowed access. That’s the key.”
Security experts have seen malicious AI assistants steal session data or private information by looking at screen content in real time. These apps often pretend to be voice assistants, and once allowed access, can silently watch user activity. When combined with AI models that create content, attackers can automate data collection, tampering, or replaying sensitive actions within apps.
“If you handle sensitive data or have regulated uses on mobile, AI Assistants are a real and present danger,” said Kai Kenan, Appdome VP of Cyber Research. “Detecting and controlling these tools is a must-have for any mobile security plan.”
Appdome’s new plugin uses user behavior to detect how malicious or unauthorized AI assistants interact with an Android or iOS app in real time. This includes official, third-party, or fake AI apps that pretend to be legitimate tools or gain extra permissions. Businesses can use Appdome to monitor AI assistant use or to detect and stop them using various options. They can also specify “Trusted AI Assistants” to ensure users only access approved AI tools.
