Lily Hay Newman Security May 13, 2026 10:00 AM WhatsApp Adds Meta AI Chats That Are Built to Be Fully Private The company says its new Incognito Chat allows you to use its AI chatbot without anyone else—including Meta—being able to access your conversations. Photograph: Anna Barclay/Getty Images Save this story Save this story WhatsApp said on Wednesday it is launching an AI chat function known as Incognito Chat that is built to allow users to converse privately with Meta AI —such that Meta itself cannot access the questions or answers. The feature is based on WhatsApp's Private Processing scheme , which debuted a year ago and already underlies WhatsApp's existing AI features, including message summarization and composition tools. The idea of Incognito Chat is to create a way for WhatsApp to offer AI chat integration that does not conflict with the communication platform's commitment to end-to-end encryption, the privacy scheme in which only direct participants in a conversation can read messages or hear a call. Most generative AI platforms now offer some type of “incognito mode,” but these features are usually designed to separate users from the questions they ask and the answers they receive rather than including a mechanism to entirely shield those questions and answers from the provider's view. With Incognito Chat, WhatsApp will only be able to see that an account used the feature, according to Meta. Meta invites third-party audits and vulnerability reports on Private Processing and says Incognito Chat itself is going to be subject to expert oversight so third parties can verify that code Meta ships for the feature is durable. But as with any cloud system, chatting with Meta AI on WhatsApp will ultimately involve trusting Meta in the same way using WhatsApp for communication does. “A big part of our work at WhatsApp relates to when there’s a perfect solution that’s very hard to use—how do you give people other options that still bring them the privacy benefits?” WhatsApp head Will Cathcart tells WIRED. “With AI, from a privacy standpoint, you’d want to run everything on your own phone, but the benefit of these models is using larger and larger compute to make them work. So the challenge is how do you build something in a data center that’s not going to fit in your pocket but has the same types of security properties. Incognito Chat is kind of like we’re running a giant phone for AI and we don’t have the passcode.” Incognito chats are ephemeral by default and will disappear once your conversation is over. Cathcart says that WhatsApp may work on developing an option within Private Processing to retain some or all of these conversations over time for users who want that history. And for now, Incognito Chat is text-only. Cathcart says support for image processing and voice recognition is in the works. Reducing latency in any way possible, including by optimizing routing, has been key to making Incognito Chat usable given the extra demands of running AI within the secure cloud environment of Private Processing. “Incognito Chat handles all AI inference in a Trusted Execution Environment that ensures your messages are not accessible to us,” Meta CEO Mark Zuckerberg wrote in a post on Wednesday. “The conversations on your phone also disappear when you exit the session. This is different from other disappearing AI products where your conversations logs often remain on other companies' servers for many months.” Meta is also adding Incognito Chat as an option in the Meta AI app. Another feature announced on Wednesday alongside Incognito Chat is “Side Chat with Meta AI,” which integrates the ability to essentially DM with Meta AI about a text chat you're having with another person or group of people. The idea is to give users a way to ask AI for restaurant suggestions or about a movie your friends are discussing without essentially leaking these locations or interests to a company. “It's sensitive information, and you should’t have to screenshot a conversation and upload it to a regular chatbot to be able to ask a question about what a friend is talking about or get an AI summary of what's been said in a big group chat,” Cathcart says. Incognito Chat includes a feature that will be on by default but can be turned off to allow its AI model to search the web in an anonymized way for up-to-date information. With more than 3 billion users around the world, Incognito Chat may offer many people their first opportunity to interact with an AI chatbot. So the fact that it is built to be privacy preserving is significant. “It's AI that runs inside Meta's hardware security modules,” says Matt Green, a Johns Hopkins cryptographer who provided input about Private Processing. He says Meta described the Incognito Chat system to him and has kept him updated on its progress. “I have confidence that if you want to talk to an AI without anyone else seeing your conversation, including Meta, this will do the job.” Green and others have pointed out previously, though, that any secure cloud-based solution simply has trade-offs because the system becomes a prominent and incredibly valuable target for attackers. And, as is already the case for financial services and other systems that require secrecy, the stakes are extremely high in a platform where users are assuming a high level of privacy and potentially sharing deeply personal information. Zuckerberg wrote on Wednesday that he is “proud that [Meta Superintelligence Labs] is the first lab to deliver private AI.” And Meta does seem to have a focus on private AI, at least for now. But the company is, of course, still a behemoth with many often competing priorities. This week, Meta entirely eliminated opt-in end-to-end encryption from Instagram Direct Messages after promising for years that it would roll out default end-to-end encryption on Instagram chat. Comments Back to top You Might Also Like How to find us: Add WIRED.com to your preferred sources in Google These women are trying to optimize their vaginas Big Story: AI gig work is the new waiting tables —and it's soul-crushing This summer, the American water crisis becomes real Event: How to adapt, compete, and win in the next era of business Lily Hay Newman is a senior writer at WIRED focused on information security, digital privacy, and hacking. She previously worked as a technology reporter at Slate, and was the staff writer for Future Tense, a publication and partnership between Slate, the New America Foundation, and Arizona State University. Her work ... Read More Senior Writer Topics privacy artificial intelligence encryption security WhatsApp Read More Thousands of Vibe-Coded Apps Expose Corporate and Personal Data on the Open Web Companies like Lovable, Base44, Replit, and Netlify use AI to let anyone build a web app in seconds—and in thousands of cases, spill highly sensitive data onto the public internet. Andy Greenberg OpenAI Rolls Out ‘Advanced’ Security Mode for At-Risk Accounts OpenAI is rolling out Advanced Account Security for people concerned that their ChatGPT or Codex accounts could be potential targets of phishing attacks. Lily Hay Newman Elon Musk’s XChat App Is More Like Facebook’s Messenger Than Signal Rather than launching a sleek, worthwhile encrypted messaging app, Musk dropped a junky, insular extension of his social media platform. Reece Rogers Update Your iPhone Now for Better Encrypted Messaging With Android As part of the iOS 26.5 update, Apple’s Messages app can now encrypt texts between some iPhone and Android smartphones. Reece Rogers Which Meta Glasses Will Shield You From the Sun and Grant You Social Media Fame? Meta is unquestionably winning the face-wearable war. Can you trust the company? Maybe not. But these are some of the nicest glasses I've ever worn. Adrienne So In the Wake of Anthropic’s Mythos, OpenAI Has a New Cybersecurity Model—and Strategy OpenAI says its safeguards “sufficiently reduce cyber risk” for now, while GPT-5.4-Cyber is a new cybersecurity-focused model. Lily Hay Newman The Top New Features in Google’s Android 17—and Gemini Intelligence—Coming This Summer You’ll soon be able to generate your own widgets or ask Gemini to finish a booking in Chrome on Android. Julian Chokkattu Hackers Hate AI Slop Even More Than You Do It's not just you. Scammers, hackers, and other cybercriminals are complaining about “AI shit” flooding platforms where they discuss cyberattacks and other illegal activity. Matt Burgess 5 AI Models Tried to Scam Me. Some of Them Were Scary Good The cyber capabilities of AI models have experts rattled. AI’s social skills may be just as dangerous. Will Knight You Can Disable Gemini in Chrome if It’s Freaking You Out Chrome users were caught off guard by a 4-GB Google AI model baked into Chrome, sparking privacy concerns. The good news: You can easily uninstall it. The bad? You might not want to. Lily Hay Newman 5 Reasons to Think Twice Before Using ChatGPT—or Any Chatbot—for Financial Advice As people increasingly rely on AI chatbots for guidance, even on financial matters, a healthy dose of skepticism is critical. Reece Rogers Google's AI Mode Update Tries to Kill Tab Hopping in Chrome Google latest update to AI Mode in its Chrome browser is designed to keep the chatbot-style search tool always around once you start an online search journey. Reece Rogers