Jatter is built around a simple principle: personalized AI should not require giving up control of your personal data.
To provide personal answers, Jatter can—with your explicit permission—learn information from websites you use, including pages inside logged-in accounts. We protect that information with client-side encryption wherever our current architecture allows, apply strict data-retention controls to service providers, and never use your data to train AI models.
We also believe trust requires being candid about what is not yet private by design. This page explains what Jatter learns, how your data is protected, where it is processed, and the limitations we are still working to address.
What does Jatter learn?
With your explicit permission, Jatter can learn information from web pages inside your logged-in accounts as you browse them.
For example, you might grant Jatter permission to learn pages from a financial, medical, school, or other personal website so that it can later help answer questions using information from those pages.
Permission is granted on a website-by-website basis. Jatter does not learn authenticated pages from a website unless you have authorized it to do so.
When does Jatter learn a page?
Jatter learns an authenticated web page only when:
- You have granted Jatter permission to learn from that website.
- You open or navigate to a page on that website while signed in.
Granting permission does not give Jatter unrestricted access to your account. Jatter learns pages as you visit them; it does not independently browse through the account on your behalf.
What is client-side encrypted?
Jatter currently uses client-side encryption for:
- Your chats after processing is complete.
- Data learned from authenticated websites after processing is complete.
Client-side encryption means this retained data is encrypted using keys controlled on your device before it is stored. Jatter’s servers do not have access to the unencrypted contents of this stored data.
However, client-side encryption does not mean the data remains encrypted during every stage of processing. Chats and authenticated web pages must currently be handled temporarily in unencrypted form so Jatter can generate responses and learn their contents.
Where third-party services are involved, we use providers and contractual safeguards intended to minimize retention and prevent secondary uses of the data.
When is data temporarily unencrypted?
Processing chat responses
Your chat must be temporarily unencrypted when it is sent to the hosted large language model used to generate a response.
The chat is sent without information intended to directly identify you. The AI provider may temporarily retain this anonymized data for security, abuse prevention, and compliance purposes.
It is not used to:
- Train AI models.
- Build an advertising profile about you.
- Track your activity across Jatter.
After the response is processed, the retained copy of the chat in Jatter is client-side encrypted.
Learning authenticated web pages
Information from an authenticated web page must currently be processed temporarily in unencrypted form before the learned data can be encrypted and stored.
The retained learned data is client-side encrypted after processing.
We are working toward moving more of this learning process onto the user’s device—particularly on desktop—so that the exposure of unencrypted authenticated data can be minimized or, where technically possible, eliminated.
Open-web reasoning
Jatter may send relevant portions of a chat to a web search service when current or open-web information is needed to answer a question.
Data sent to this service is covered by a zero-data-retention policy. It is not retained by the service after processing.
What is not currently client-side encrypted?
The following data is not currently protected with client-side encryption:
- Browsing history.
- User notes.
- User profile and account data.
- Chats while they are temporarily being processed.
- Authenticated website data while it is temporarily being processed.
Encrypting browsing history and user notes is on our roadmap.
Some user profile information must remain accessible to Jatter’s systems so we can operate your account, provide the service, manage subscriptions, and support account recovery and deletion.
What data does Jatter retain?
Depending on how you use the product, Jatter may retain:
- Client-side encrypted chats.
- Client-side encrypted data learned from authenticated websites.
- Browsing history.
- User notes.
- User profile and account data.
We retain this information to provide features such as conversation history, personalized answers, notes, browsing functionality, account management, and synchronization across supported devices.
What data is sent to AI and search providers?
Large language model provider
Jatter sends anonymized, temporarily unencrypted chat content to the hosted large language model used to generate responses.
The provider may temporarily retain this data for security and compliance purposes. It may not use Jatter user data to train its models or track individual Jatter users.
When information learned from an authenticated website is needed to answer a question, relevant portions of that information may also be included in the request sent to the model.
Web search provider
When a question requires open-web research or reasoning, relevant chat content may be sent to our web search provider.
All Jatter user data sent to this service is subject to a zero-data-retention policy.
Does Jatter train AI models on user data?
Never.
Jatter does not use your chats, browsing activity, notes, learned website data, or profile information to train AI models.
We also require the AI services used by Jatter not to use Jatter user data for model training.
Does Jatter sell user data?
No.
Jatter does not sell your personal data.
We do not operate an advertising business based on tracking your browsing behavior, and we do not share your personal information with advertisers for targeted advertising.
Can users delete their data?
Yes.
You can delete:
- Individual chats.
- Data learned from authenticated websites.
- Browsing history.
You can also delete your Jatter account. When an account is deleted, Jatter deletes the user data associated with that account, subject to any limited information we may be legally required to retain.
Because some third-party providers may temporarily retain anonymized data for security and compliance purposes, deletion from Jatter may not immediately remove data held within a provider’s limited compliance-retention window.
What are Jatter’s current privacy limitations?
Jatter is designed to improve privacy beyond the conventional model of AI personalization, but the system is not yet private in every respect.
Browsing history is not yet encrypted
Your browsing history is currently stored without client-side encryption. Client-side encryption for browsing history is on our roadmap.
Browsing-history collection cannot yet be disabled
Jatter does not currently provide a setting to disable the collection of browsing history while using the browser. We plan to add more user control over this behavior.
Notes are not yet encrypted
User notes are not currently protected with client-side encryption. Encrypting notes is on our roadmap.
Authenticated website learning requires temporary unencrypted processing
Although retained learned data is client-side encrypted, authenticated web pages must currently be processed temporarily in unencrypted form.
We plan to move more of this processing onto the user’s device over time—at least on desktop, and potentially on mobile—so that the exposure of authenticated data can be reduced or eliminated where possible.
Chats require temporary unencrypted processing
Chats must currently be temporarily unencrypted to use a hosted large language model.
The AI provider may temporarily retain anonymized chats for security and compliance purposes. We are working toward placing this processing under a full zero-data-retention policy in the future.
Our direction
Privacy is not a feature we consider finished. It is an architectural direction.
Our goal is to steadily move more personal-data processing onto the user’s device, expand client-side encryption to additional categories of data, reduce the amount of information exposed during processing, and give users more precise control over what Jatter learns.
We will continue updating this page as Jatter’s privacy architecture evolves.