False WhatsApp chain letter warns of alleged AI access to chats

What the "Advanced Chat Privacy" function actually does

The AI function Meta AI, which has been gradually introduced since 2025, processes – according to WhatsApp's privacy notices regarding Meta AI – exclusively content that users actively send to the AI – for example, via an @ mention, direct questions, or AI functions such as summaries. Only then does this data leave the end-to-end-encrypted path and get processed in plain text on Meta servers. WhatsApp messages in one-on-one and group chats remain end-to-end encrypted by default. Neither Meta nor Meta AI can read these contents.

WhatsApp emphasizes that chats are private regardless of the "Advanced Chat Privacy" setting. The "Advanced Chat Privacy" function mentioned in the chain letter does actually exist. It can be activated per chat by tapping the group name at the top of a group chat. It blocks the export of chat history, prevents the automatic saving of media on participants' devices, and prevents the use of Meta AI functions in the respective chat. When activated, a system message appears informing all participants.

However, according to the MDR report, the option primarily protects against privacy violations by other users in the chat, not against Meta itself. Those who activate the function prevent members from exporting the history or automatically saving photos and videos in the gallery – but not Meta using its own AI tools within the chat. This distinction is missing in the chain letter, which presents the toggle as a panacea against an allegedly impending AI surveillance.

Metadata: the actual privacy problem

Despite the end-to-end encryption of content, WhatsApp collects extensive metadata regardless of Meta AI and regardless of Advanced Chat Privacy. This includes phone number, device information, IP address, timestamps of messages and calls, contact lists, online status, and location data. WhatsApp has been automatically evaluating all of this metadata for years.

This metadata is not end-to-end encrypted and can be shared with other Meta services – supported by the legally contested construct of "legitimate interest" under the GDPR. This metadata is then traded on global data brokers. Specialized firms create profiles about each of us from it. These are then bought by recruitment agencies, insurance companies, credit institutions, intelligence services (including German state criminal police offices and the FBI), and many others.

What users can do now

The chain letter also calls on its recipients to spread the warning. Fact-checkers, however, advise against doing exactly that: anyone who receives a dubious message should not share it without verification, but should first check the source – in this specific case, for example, via Mimikama or WhatsApp's official privacy notices. AI functions in WhatsApp, such as creating summaries, are only activated, according to WhatsApp, when users explicitly request them.

As alternatives to WhatsApp, the MDR report names Signal, Threema, or Wire, which are also now very widely used and collect significantly less metadata. Those who nevertheless stick with WhatsApp can use the "Advanced Chat Privacy" function to at least restrict the export of history and automatic media storage in groups – however, according to WhatsApp and Mimikama, there is no automatic reading by an AI.

The article was published by MDR AKTUELL. The central message: the threat described in the chain letter does not exist in the form presented; the activated toggle does not change the fundamental encryption of the chats. However, the actual discussion about metadata and its trade is not affected by this.