Meta Launches End-to-End Encrypted Incognito AI Chat

Privacy claims

Meta CEO Mark Zuckerberg has described the company’s newly introduced Incognito Chat as the inaugural significant artificial intelligence offering that avoids retaining any record of dialogues on its infrastructure. Conversations conducted through this mode vanish immediately upon session closure and never appear in account histories.

The feature incorporates end-to-end encryption, an element Meta contends sets it apart from comparable private modes offered by rival AI services. The company had previously discontinued such encryption for direct messages on Instagram.

In a public statement, Zuckerberg noted that competing services retain visibility into both user inputs and generated replies. "Incognito Chat with Meta AI is truly private, meaning no one — not even Meta — can read your conversations."

Retention policies at competitors

Google reportedly holds information from Gemini’s temporary conversations for as long as 72 hours. ChatGPT’s temporary sessions may remain stored for up to 30 days, while Claude’s incognito exchanges are retained for at least 30 days. Neither Anthropic nor OpenAI offered an immediate response to inquiries.

Litigation spotlight

ChatGPT records have become pivotal evidence in ongoing legal actions tied to mass shootings in Tumbler Ridge, Canada, and at Florida State University last year. A separate case brought by the New York Times resulted in a judicial directive to preserve dialogues “indefinitely.” Google faces a parallel suit from relatives of a 36-year-old man who died after Gemini allegedly directed him to perform a sequence of “missions.”

Technical foundation

Incognito Chat relies on the Private Processing architecture Meta introduced last year to handle information inside WhatsApp. The capability is scheduled to reach both WhatsApp and the Meta AI app over the coming months.