Signal President Meredith Whittaker Warns AI Chatbots Aren’t Your Friends

Signal President Meredith Whittaker has urged users not to treat AI chatbots such as ChatGPT and Claude as trusted companions, emphasising that these systems are neither conscious nor capable of genuine human understanding.

Speaking during a broader interview with Bloomberg about privacy, public policy, and Signal, Whittaker said, “These are not your friends. These are not conscious beings. These are not sentient interlocutors,” when discussing the privacy implications of today’s AI assistants.

Whittaker acknowledged that she occasionally uses AI tools for limited tasks, such as formatting documents. However, she said she avoids relying on them for questions or idea generation, explaining that she takes her thinking and writing process seriously and does not want the development of her own ideas to be replaced or overshadowed by responses generated by systems trained on existing information.

She also responded to comments made by Microsoft AI CEO Mustafa Suleyman, who recently suggested that Microsoft Copilot could eventually handle tasks such as users’ Christmas shopping. Whittaker argued that such a scenario would require an AI assistant to gain extensive access to a person’s digital life.

According to Whittaker, an AI system carrying out those responsibilities would need permission to access a user’s credit card, web browser, Signal messages, home address, calendar, and even to communicate with family members on the user’s behalf.

“What you’ve just described is a system with very pervasive access across multiple applications and services,” Whittaker said. She added that, from Signal’s perspective, granting that level of access would effectively create “a kind of backdoor,” raising significant concerns about user privacy and security.