Is WhatsApp’s unread message summary really private?

Meta’s latest feature for WhatsApp has sparked questions about privacy in the age of AI. The new “unread message summary” lets Meta’s AI create recaps of group chats we haven’t opened, helping us catch up without scrolling through hundreds of messages. But WhatsApp’s long-standing promise of end-to-end encryption raises an obvious concern: how can an AI read our messages without breaking encryption?

A breakthrough in encrypted AI processing

Meta has introduced a technology called “Private Processing” that allows AI to summarize chats without accessing message content in the traditional sense. The process happens inside a sealed virtual machine, a kind of secure hardware chamber where messages are decrypted and summarized — but then instantly deleted.

The messages never leave the encrypted environment and are never stored. If anyone attempts to intercept the process, the system automatically shuts down, preserving privacy. Even Meta itself can’t read the messages, and summaries are not linked to user identity or used for ads.

Balancing usefulness and trust

Still, this innovation arrives at a time when Meta’s track record on privacy remains under scrutiny. From Cambridge Analytica to past data leaks, the company’s promises don’t always inspire confidence. Experts warn that today’s optional summaries could open the door to future features like AI recommendations or ads, even if anonymized.

Setting the tone for private AI

WhatsApp’s move could set a standard for AI use in encrypted environments. Apple has already launched Private Cloud Compute, and other platforms like Signal and Telegram may follow. Meta’s solution shows how advanced AI features can coexist with strong privacy—at least in theory.