Meta has unveiled a new feature for WhatsApp designed to provide users with a private means of interacting with Meta AI. Dubbed “Private Processing,” this option will be available in the coming weeks and is characterized by its privacy protections, ensuring that Meta, WhatsApp, and any third-party entities will not have access to the exchanges made while using it.
According to Meta, the Private Processing feature allows users to request AI-driven tasks, such as generating chat summaries. Notably, the system ensures that it does not maintain access to user messages once the session concludes, which is aimed at thwarting any potential post-session access by hackers, the company stated.
Meta intends to minimize the risk of attacks that target users without a comprehensive system breach. Furthermore, the company has expressed a commitment to enabling third-party auditors to verify the privacy and security promises associated with Private Processing. To strengthen its security framework, this feature is now included in Meta’s bug bounty program, with plans for a detailed security engineering design paper to be published prior to the feature’s public rollout.
The operational structure of Private Processing bears similarities to Apple’s Private Cloud Compute (PCC). Like Apple, Meta plans to route Private Processing requests through an external provider using OHTTP, a protocol designed to mask the users’ IP addresses. However, as noted by Wired, a significant distinction lies in the fact that all AI requests on WhatsApp are processed via Meta’s servers, and the feature must be actively initiated by the users. In contrast, Apple typically prioritizes on-device AI processing, resorting to PCC only when requests necessitate server intervention.
