WhatsApp function enables AI deactivation in chats – but is not necessary

The Meta AI chatbot has been integrated into WhatsApp since March. The AI, which can be found prominently in the app as a blue ring on the start page, cannot be removed, even though this question is apparently asked again and again. However, with the help of the "Advanced Chat Privacy" function, which was introduced in April, users can block the AI from individual and group chats so that chat participants cannot activate the AI.

However, the fact that Meta AI can access all data in WhatsApp, as rumored in recent chain letters, is not correct – and certainly not without the user's consent.

Meta AI in WhatsApp can only be activated by the user

The AI in WhatsApp can only access the user's data if the user interacts with it and instructs it to do so. All individual and group chats are end-to-end encrypted by default so that only the participants in a chat round have access to the information shared in it.

The AI functions integrated into WhatsApp only take effect when users actively use them. This means that there is no automatic access to content, let alone phone numbers. As a reminder, users actively provide the latter to Meta when they synchronize their address book with WhatsApp to find their contacts.

Only when Meta AI is addressed via the @Meta command in a chat window does the user open the door to the chat for Meta's AI. The AI in the chat can be used to summarize missed messages (initially only in the USA) – This function is not yet available in Germany.

As the company explains, a session with a cloud server is established for this private AI summary – anonymized by Oblivious HTTP (OHTTP) via third-party intermediate stations. According to the explanation, the session is encrypted with a temporary key that "Meta and WhatsApp do not have access to"; only the end device and the private processing servers can use it: Meta emphasizes that "no one other than the user's device or the selected TEEs (Trusted Execution Environment)" can decrypt the request.

In addition, the "content and detailed circumstances of chats between users of telecommunications services", which include WhatsApp, are "subject to the special protection of telecommunications secrecy (Section 3 TDDDG), which is enshrined in the German Basic Law (Article 10 GG) and the violation of which is punishable by law (Section 206 of the German Criminal Code)," erklärt the Federal Commissioner for Data Protection and Freedom of Information (BfDI).

Enhanced data protection blocks users from using AI in chats

For even more control over the meta AI, although not really necessary, users can activate the "Extended data protection" function. However, this feature must be activated individually for each chat. This is done via the respective settings, which can be accessed by pressing the recipient of the chat or the group name (not the three-dot menu). If you now type @Meta in a chat window, you will be informed that the function is not available in the respective chat.

Activating the "Extended data protection function" not only blocks Meta AI access in the chat, but more: the feature simultaneously blocks the export function for the respective chat and media access, so that images and co. are not saved in the respective media gallery of the device. As mentioned, this function is not required to prevent Meta AI from unauthorized data access. It is sufficient if none of the participants in a WhatsApp group chat use it.

According to Meta, the "Extended Privacy" function is primarily used in WhatsApp groups in which someone does not know all participants in detail, but in which more confidential matters are also discussed. The company cites conversations about health problems in support groups as an example.

(afl)