An AI keyboard is useful because it sits where you write. That is also why you should evaluate it more carefully than an ordinary app. This checklist helps you ask concrete questions before enabling an Android keyboard with AI features.
It is a companion to the longer Android keyboard privacy guide. Use it for Synapse, Gboard, or any other keyboard—the point is to compare actual behavior and documentation, not to trust a logo or a generic privacy promise.
1. What text is processed, and when?
Look for a plain explanation of what stays on the device and what is sent to a server. “Private” is not enough. A useful policy explains the trigger.
For an AI keyboard, the important distinction is between normal typing and an explicit AI action. Synapse is designed so normal typing stays on-device and text is processed only when you select it and deliberately run a prompt. If another keyboard sends context continuously for suggestions or cloud features, its policy should say that clearly.
2. What does the app retain?
Ask whether the service stores prompt text, generated output, usage records, saved prompts, or account information. Retention is not automatically bad—history can be a feature—but you should know where it lives, how long it remains, and how to request deletion.
Read the policy’s account-deletion process before you need it. If the answer is vague, do not assume that deleting the app deletes server-side data.
3. Which permissions are requested?
Open Android Settings, find the keyboard, and inspect its permissions. A keyboard may need internet access for cloud AI and vibration for haptic feedback. Permissions such as contacts, microphone, SMS, location, or storage deserve a specific explanation tied to a feature you actually want.
Also review which keyboards are enabled. Disable keyboard apps you do not use, particularly old ones you installed for a one-time test.
4. What happens in sensitive fields?
Password and one-time-code fields are special cases. Android marks many of these as sensitive input, and responsible keyboards should not learn from or retain the text. A stronger habit is to use a password manager’s autofill feature, which avoids typing a credential through the keyboard at all.
Never paste passwords, card numbers, authentication codes, recovery phrases, or confidential client data into an AI prompt. Even when a tool has a careful privacy design, the safest sensitive text is the text you do not submit to a third-party service.
5. Can you understand the business model?
Know how the keyboard is funded. A visible price, subscription, credits, or a documented company relationship gives you something to evaluate. If a keyboard promises unlimited cloud AI for free with no explanation, look more closely at its privacy policy, data practices, and account requirements.
Synapse offers a free download and free credits, then uses pay-as-you-go top-ups for AI use. Review the privacy policy and purchase terms for the current product terms instead of relying on an old blog post.
6. Is the developer and support path real?
Check whether the app has a reachable website, a privacy policy, clear support contact information, and a documented way to report a problem or request account deletion. These do not prove that software is safe, but an accountable product should make the basic information easy to find.
For Synapse, product information is available on the features page, and support is available at the address listed on the About page.
A final decision rule
Enable an AI keyboard only when you can answer these questions in plain language:
- What happens to normal typing?
- What happens when I run an AI prompt?
- What data is retained and how can I remove it?
- Which permissions are present and why?
- How can I reach the developer if something goes wrong?
If you cannot answer them after reading the product pages and policy, choose a different keyboard or keep AI features disabled until you can. Convenience is useful; clarity about your data is non-negotiable.