Is It Safe to Let AI Summarize Your Messages?

AI summaries are one of the most useful things to happen to messaging in years. Instead of reading a hundred messages to find the five that matter, you read a short briefing.

But there's an obvious catch: to summarize your messages, the AI has to read them. And your messages hold some of the most private things in your life: health, money, family, work secrets.

So is it safe? The honest answer is: it depends on the app, and you deserve to know exactly how each one works. Here are the questions worth asking, and how we answer them for YAFOY's Insight.

1. Where does the AI run? Some AI features run entirely on your phone. Others send text to a cloud AI model. Cloud models are usually more capable; on-device is usually more private.

2. What exactly is sent? Your whole message history, or only recent excerpts? Full messages, or trimmed snippets?

3. Is anything removed first? Does the app strip out sensitive details like phone numbers, email addresses, or bank and login codes before sending?

4. What's stored, and for how long? Are the inputs or summaries saved on the company's servers? For how long?

5. Is your data used to train AI? Check whether the app or its AI provider uses your content to improve their models.

6. Can you turn it off and delete it? You should be able to stop using the AI feature, and delete your data entirely.

If an app can't answer these clearly, that's an answer too.

We'd rather you know exactly what happens than take our word for "secure."

• Where it runs: Insight uses a cloud AI model (Google Gemini, accessed through our platform provider). It does not run entirely on your phone or desktop computer.
• What's sent: short excerpts from recent conversations, not your full history. Up to 25 conversations, with excerpts of up to 300 characters each, plus context like the sender and time.
• What's removed first: email addresses, phone-like numbers and some reference numbers are masked. Messages that look like banking or one-time codes are left out, and Signal conversations are excluded by default. This redaction is partial: names and other details in the text can remain, and dates and amounts are kept on purpose so Insight can spot deadlines and payments.
• What's stored: your communication inbox is stored on the device you use, including your desktop browser. It is not synced from your phone to your computer. Saved Add captures are account-backed, and established temporary delivery queues and Insight processing still involve servers. Insight's generated summaries are saved to your account [add the retention period confirmed in the corrected privacy policy].
• Training: [state the confirmed answer for YAFOY and its AI provider before publishing].
• Control: you can delete your account and data from Settings. [Confirm once full deletion is verified.]

You can read the full details on our privacy page.

• Keep the most sensitive conversations out. If an app lets you exclude certain apps or chats from AI, use it for things like medical or legal conversations.
• Don't paste passwords or codes into chats. That's good advice with or without AI.
• Check the original before you act. AI summaries can misread or miss things. Treat them as a starting point.

AI summaries are safe enough for many people when the app is honest about what it sends, removes and keeps, and lets you leave. Those are the standards we hold Insight to, and the ones you should expect from any app that reads your messages. See how Insight works →

Gmail, Outlook, Slack and Teams connect directly in desktop YAFOY without reading phone notifications. The local desktop inbox and the local phone inbox are separate. When Insight is generated, recent excerpts from the device in use still go through partial redaction before cloud AI processing. A browser window does not make that analysis offline or anonymous. Saving an item with Add is different: those captures are account-backed for access through Memory on other devices.