WhatsApp launches 'private processing' mode for AI chatbot conversations

WhatsApp has launched a new 'private processing' mode that keeps user prompts to its Meta AI chatbot off the company's main servers, the BBC reports. The firm says messages are decrypted only inside hardware-secured enclaves and deleted once a response is generated. The feature is rolling out first in English before expanding to other languages.

Security researchers who spoke to the BBC said keeping chatbot inputs out of advertising or training pipelines is a meaningful privacy gain. Others cautioned that the claims need independent verification, and Meta says it will publish a technical paper describing the architecture. The rollout is opt-in rather than default.

The announcement comes as AI assistants face mounting regulatory pressure. The European Union and the United Kingdom are scrutinising whether personal data is being used to train assistant models without sufficient consent. With more than two billion users globally, WhatsApp's design choice could set a benchmark for rivals.