Signal's Meredith Whittaker warns AI chatbots are 'not your friends' and calls agentic AI a backdoor to privacy

Signal president Meredith Whittaker warned this week that AI chatbots "are not your friends" and called the agentic AI vision promoted by Microsoft a backdoor that undermines end-to-end encryption. Her comments came in a Bloomberg interview published June 20.

Signal's encryption protocol is used by WhatsApp, which counts more than two billion users. That reach gives Whittaker's privacy warnings global weight as tech giants race to embed AI agents into their operating systems and apps.

Whittaker's critique of agentic AI

Whittaker took direct aim at Microsoft AI CEO Mustafa Suleyman's prediction that users would let Copilot handle all their Christmas shopping by eavesdropping on family group chats. She methodically listed the access such a system would require: "my credit card, my browser, my Signal, the ability to message my siblings on my behalf, my home address, my calendar." She called that level of access a backdoor.

Her interview also pushed back on the growing tendency to anthropomorphize chatbots. Key points from her remarks:

• AI systems are not conscious beings or sentient interlocutors.
• She uses AI only to format documents, never to ask questions or work through ideas.
• Agentic AI systems are structurally incompatible with end-to-end encryption because they need to read plaintext messages before encryption or after decryption.
• She identified prompt injection as the likeliest first exploit path against encrypted messaging platforms.

Broader implications for privacy and security

Whittaker has been making this case with increasing urgency. In January 2026 she warned at Davos that agentic AI was perilous for secure applications, and in an essay for The Economist she accused operating system vendors of hollowing out Signal's ability to guarantee privacy by embedding agents into platforms. She has previously stated that Signal would leave the EU rather than comply with any law requiring it to compromise its encryption.

Microsoft, Google, Apple, and OpenAI are all building agent-first architectures. Whittaker's argument is that productivity gains from surrendering control of messages, calendars, contacts, and financial data to a corporate AI system are not gains but a transaction in which users trade privacy for convenience without understanding the terms. As these systems become more pervasive, the risk that hackers or governments will target the rich databases they create grows exponentially.