Signal Foundation president Meredith Whittaker said AI chatbots like ChatGPT and Claude "are not your friends" and are neither conscious nor sentient interlocutors, renewing her warning about the privacy risks posed by "agentic AI" that carries out tasks autonomously.
Signal Foundation · Privacy Warning
"AI Chatbots Are Not Your Friends"
Signal president Meredith Whittaker warns that autonomous AI agents demand deep OS-level access — breaking the "blood-brain barrier" between system and apps, and effectively neutralizing the encryption that protects your messages.
~70M
Signal monthly active users protected by end-to-end encryption
5+
Data sources an agent must reach: calendar, browser, card, contacts, messages
2023→26
Years Whittaker has warned AI is "fundamentally a surveillance technology"
How the protection collapses
An AI agent reading data continuously at the OS level sits beneath the app — where encryption no longer protects you.
App layer · protected
Signal messages secured by E2EE — content unreadable.
↓
"Blood-brain barrier"
The OS / app separation that keeps the system out of your messages.
↓
Agent at OS level
Reads everything continuously — encryption effectively bypassed.
Three structural concerns
Scope of access
Broad permissions to calendar, browser, credit card, contacts and messages.
Architecture
OS/app-layer separation collapses, circumventing encryption protections.
Legal risk
Aggregated data could become subject to subpoenas or breaches.
THE BOTTOM LINE
"An agent is only as powerful as the access you give it."
This is not a critique of one product or benchmark, but of agentic AI's broader rollout — where autonomous agents, device scanning and digital advertising converge into a new surveillance architecture.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…