WhatsApp AI is now available across Meta’s range of apps, whether it’s WhatsApp or Instagram. However, one area where the company hasn’t yet introduced advanced AI is in WhatsApp chats. There are no message summaries or writing suggestions, mainly because the chats are encrypted to protect user privacy and security.
That said, the company is now promising a way that will bring AI capabilities like message summarisation and writing assistance, while still maintaining its strong privacy protections. This is made possible by a new technology called ‘Private Processing’.
In short, WhatsApp will soon offer AI-powered tools such as message summaries without compromising user privacy, according to Meta. The company explained all of this in a detailed post, adding that the feature will be optional, transparent, and give users more control. Here, we simplify how this new technology works, what failsafes exist, and the structure behind threat detection.
What is Private Processing?
Currently, most AI features rely on Large Language Models that run on servers rather than on a user’s mobile device. Naturally, this makes the data visible to external systems, which is a concern when handling private information such as WhatsApp chats. This is where technologies like Private Processing come into play.
Meta explains that Private Processing is a technology it has developed to enable the use of AI in a secure and privacy-preserving manner. It uses secure ways to allow AI to assist with tasks, such as message summarisation, without compromising privacy or breaking end-to-end encryption.
Meta says that “No one except you and the people you’re talking to can access or share your personal messages, not even Meta or WhatsApp.”
Also Read: Google Maps summer travel tips: 5 essentials you must keep in mind
Maintaining User Privacy While Handling AI Requests Online: How Is It Possible?
AI and privacy: that’s a combination that is often met with scrutiny, and not usually associated with privacy. Time and again, both users and experts have raised concerns about how generative AI is trained and the significant privacy risks it poses. So, in this context, how can Meta’s new technology enable AI features without compromising user privacy?
According to Meta, the feature has been designed with several “foundational fundamentals” in mind: Confidential Processing, Enforceable Guarantees, and Verifiable Transparency.
Put simply, Meta claims that Private Processing prevents any other system, including Meta itself, WhatsApp, or any third party, from accessing user data. Meta is also allowing both users and security researchers to audit the behaviour of Private Processing, ensuring its privacy and security guarantees can be verified.
In addition, Meta is building in fail-safes to ensure that even if a bad actor tries to target a user, they would need to take down the entire Private Processing system to succeed. Furthermore, Private Processing must not retain access to user messages once the session has ended.
Meta Acknowledges That Issues Could Arise
Meta knows that threats can affect this experience and technology at large, and that is why its threat model includes three key components. These include identifying threat actors, which may include groups or individuals that could attempt to compromise Meta’s security. Also, it is identifying the ways Meta’s assets could be compromised, including techniques hackers might use. Simply put, Meta is making sure that vulnerabilities that come with implementing generative AI features are properly dealt with.
MOBILE FINDER: iPhone 16 LATEST Specs, Price And More
Leave a Reply