Meta launched Incognito Chat with end-to-end encryption for AI

Meta launched Incognito Chat, a private mode for its AI assistant. Conversations are fully end-to-end encrypted and are not stored on servers. Even Meta itself cannot read them. That is the key difference from competitors' incognito modes, which still see users' questions and answers.

AI-processed from The Verge; edited by Hamidun News
Meta launched Incognito Chat with end-to-end encryption for AI
Source: The Verge. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceThe Verge ↗
Editor-in-Chief
Key metric #1Analyzing answer quality
Key metric #2Training and fine-tuning models
Key metric #3Checking for policy violations

Meta has introduced Incognito Chat — a new mode for its AI assistant where all conversations are fully private and encrypted end-to-end. According to Mark Zuckerberg, this is the first major AI product where conversations are not logged on servers at all — neither for analytics, nor for debugging, nor for anything else.

How Encryption Works

Incognito Chat uses end-to-end encryption — a technology that encrypts your messages from the moment you hit send. The text is encrypted on your device and decrypted only in the AI processor. Meta's servers store only the encrypted text, which the company technically cannot read.

This is not simply deleting history from the interface. At the infrastructure level, conversations are not saved at all — neither as logs, nor for model training, nor for toxicity checking. After the conversation ends, the data is deleted.

Difference from Competitors (and from Meta Previously)

Most AI assistants offer incognito modes that work quite differently. OpenAI (ChatGPT), Google (Gemini), Anthropic (Claude) — all allow you to disable history saving in your account. But at the company level, they still see all questions and answers for:

  • Analyzing answer quality
  • Training and fine-tuning models
  • Debugging bugs
  • Checking for policy violations

This distinction matters to Zuckerberg. In a recent post, he wrote:

"Other apps have introduced incognito-style modes, but they can still see the questions coming in and the answers going out.

Incognito Chat with Meta AI is truly private, meaning no one — not even Meta — can read your conversations."

Paradox: Meta — a company known for collecting user data — now promises complete denial of access even to others' questions.

A Shift in Meta's Policy

This looks ironic against recent history. A few months ago, Meta removed end-to-end encryption from regular messages in Instagram DMs and Messenger. Officially — to combat crime and protect children. Privacy activists cried out in horror. Now the company is adding E2E encryption to its AI assistant.

The signal is mixed: Meta realizes that people need privacy, but it's unclear what kind of privacy this is from the company itself versus from the outside.

What This Means

Incognito Chat could become a tool that pushes competitors toward more transparent policies. If users start choosing Incognito Chat not because of the UI, but because of privacy, pressure on OpenAI and Google will increase.

For users, it's simple: if you're worried that conversation data will be used to train models or sold to advertisers, there's a button to enable private mode.

Story

Key Questions Answered

How Encryption Works?

Incognito Chat uses end-to-end encryption — a technology that encrypts your messages from the moment you hit send. The text is encrypted on your device and decrypted only in the AI processor. Meta's servers store only the encrypted text, which the company technically cannot read. This is not simply deleting history from the interface. At the infrastructure level, conversations are not saved at all — neither as logs, nor for model training, nor for toxicity checking. After the conversation ends, the data is deleted.

What are the key details on Difference from Competitors (and from Meta Previously)?

Most AI assistants offer incognito modes that work quite differently. OpenAI (ChatGPT), Google (Gemini), Anthropic (Claude) — all allow you to disable history saving in your account. But at the company level, they still see all questions and answers for: Analyzing answer quality Training and fine-tuning models Debugging bugs Checking for policy violations This distinction matters to Zuckerberg. In a recent post, he wrote: "Other apps have introduced incognito-style modes, but they can still see the questions coming in and the answers going out. Incognito Chat with Meta AI i

What are the key details on A Shift in Meta's Policy?

This looks ironic against recent history. A few months ago, Meta removed end-to-end encryption from regular messages in Instagram DMs and Messenger. Officially — to combat crime and protect children. Privacy activists cried out in horror. Now the company is adding E2E encryption to its AI assistant. The signal is mixed: Meta realizes that people need privacy, but it's unclear what kind of privacy this is from the company itself versus from the outside.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…