Draft. Awaiting review by a lawyer.
The Hebrew version of this document prevails over the translation. The translation is a draft awaiting review.
Privacy policy
This policy explains what information the Commika service collects, why, who receives it, how long it is kept and what your rights are. It is written in accordance with the Privacy Protection Law, 5741-1981, including Amendment 13 to the Law, which came into force in August 2025, and the Privacy Protection Regulations (Data Security), 5777-2017. The service is still in development: at present the site, the waitlist and the demo are operating. The server and the apps have not launched yet, so the sections that describe them (what data will be stored, security, retention and providers) describe the plan, and will apply from the launch. Anything marked “planned” is an intention and not an existing state.
1. Who we are and what our role is
The Commika service is operated by [[להשלמה: שם הישות המשפטית]] (number [[להשלמה: מספר חברה או עוסק מורשה]]), [[להשלמה: כתובת]]. Hereinafter: “we”.
The division of roles between us and you:
- For your account details (name and email address), we are the database owner.
- For the coded commission data of your agency, the agent or agency is the database owner. We hold and process it on their behalf, solely in order to provide the service.
[[להשלמה: אישור עורך דין לחלוקת התפקידים בין בעל מאגר, מחזיק ומעבד]]
2. The principle: what stays on your computer and what leaves it
The system is built so that identifying details of clients never reach us. The data is divided into three levels, and only two of them leave your computer:
| Level | What is included | Where it goes |
|---|---|---|
| Level 0 | The original report files. Client names, Israeli ID numbers, addresses, phone numbers, email addresses, dates of birth, bank details, medical information and free-text columns. Full policy and account numbers and employer names. | Stays on your computer only. Not sent to the server, not sent to the AI, not written to logs and not included in crash reports. |
| Level 1 | Coded client and policy identifiers, for example “Client 17” and a coded policy ID. The link between the code and the real name or number is kept only on your devices. | Sent to our server and to the AI service. |
| Level 2 | Commission amounts, commission rates, premium or balance, product type, commission type (volume commission (heikef) or trailing commission (nifraim)), policy status, report month and policy start month, insurer code and your agent number, counts and total amounts, and finding types. | Sent to our server and to the AI service. |
- Only fields explicitly identified as allowed leave the computer. Every other column in the report is dropped by default.
- The coding is done on your computer. The coding key is kept only on your devices and does not reach us, so we cannot decode the codes.
- The server, which has not launched yet, is planned to re-check every piece of data that reaches it, and to reject it if it looks like an Israeli ID number, an email address or a phone number, or if it includes a field that was not defined in advance. The check looks for ID numbers, phone numbers and email addresses. It cannot detect names, so client names are replaced with codes on your computer, before anything is sent.
- Letters that include real policy numbers are filled in on your device only.
3. What we collect
Account details
When you sign in with Google, Apple or Microsoft, we receive from the provider your name, your email address and an internal identifier of the account at that provider. There are no passwords. We also store the agency name, the sign-in provider and your role in the agency (owner, manager or agent).
Agency profile
The agency name, the agent’s name, your agent numbers with the insurers, a phone number and email address that will appear in letters, a logo and a signature line. This is business data and not client data, and it is stored on the server so that letters are filled in automatically on all your devices. The signature line, the titles of calendar events and the subjects of letters that you write go through the same check for Israeli ID numbers, phone numbers and email addresses. The setting of whether the AI is active in the agency is also stored.
Coded data (level 1 and level 2)
The results of the checks that your computer sends to the server: check runs by insurer and month, and findings (the finding type, coded client and policy identifiers, amounts, commission rates, product and status), together with the AI’s explanations and the results of the second checker’s review. We also store draft letters (subject and body with placeholders) and calendar events and reminders you created.
Device and usage data, without content
The device type (Windows, Mac, iOS or Android), the time of last activity, and, if you approved notifications, an identifier that allows notifications to be sent to the device. The server logs record only a request ID, a user ID, the address requested, the response code and the processing time. An audit log is also kept of actions such as sign-in, uploading a run, deletion and sending a letter. None of these includes the content of requests or responses.
Google Calendar, if you connect it
Connecting Google Calendar is entirely your choice. If you connect it, we will store encrypted access tokens and the identifiers of the events created.
4. What we do not collect
We do not collect, receive or store:
- client names;
- Israeli ID numbers;
- clients’ addresses, phone numbers, email addresses and dates of birth;
- bank details and medical information;
- the original report files;
- full policy and account numbers, and employer names;
- the content of free-text columns.
If such information is sent to the server by mistake, the system is designed to reject it and not store it. To help it do so, do not enter identifying details of clients into fields you fill in manually, such as the subject of a letter or the title of a reminder.
5. Why we use the information
- to provide the service: report comparison, analysis, letters, notifications and calendar;
- to secure the service and prevent misuse;
- to support you and respond to inquiries;
- to fix faults and improve the service on the basis of usage data, without content;
- to comply with legal requirements.
We do not sell information and do not use it for advertising. The site and the apps do not include third-party tracking or analytics tools. Providing the account details is not required by law, but without them an account cannot be opened and the service cannot be used.
6. AI processing by Anthropic
To recognize the report structure, analyze the findings, check them and draft letters, we use the AI service of Anthropic, our sub-processor. Column recognition already works in the demo. The analysis, the second check and the drafting of letters will run on a server that has not launched yet, and the table describes how they are planned to work. The AI service receives coded data only:
| Task | What Anthropic receives |
|---|---|
| Column recognition | Only header rows are sent as they are: short rows, with no digits and no @ sign, where most cells are familiar column names such as “policy number” or “amount”. Every other row is masked: letters are replaced with a uniform letter, digits with #, and amounts by order of magnitude only. |
| Analysis and checking | Level 1 and level 2 data only. |
| Letter drafting | Summaries only: counts and total amounts. |
- We do not send report files. Rows that are not header rows are sent only masked. Names cannot be detected automatically, so if a client’s name appears in a header row, it will be sent, and you should remove it from the file before uploading it.
- You can switch the AI off in the agency settings. When it is off, no request is sent to Anthropic, and the columns are chosen manually.
- Calculations of differences, amounts and thresholds are done in our code and not by the AI.
- The AI can make mistakes. Findings are gaps to check and not proof of an error (see the terms of use).
- Retention and use of data at Anthropic: [[להשלמה: תנאי השמירה והשימוש בנתונים אצל Anthropic לפי ההסכם בפועל, והאם הופעלה אי־שמירה]]
- The processing at Anthropic may take place outside Israel. Such a transfer is made in accordance with the Privacy Protection (Transfer of Data to Databases Abroad) Regulations, 5761-2001, and subject to the provider’s undertaking to protect the information. [[להשלמה: בסיס ההעברה מחוץ לישראל, אם נדרש]]
8. Storage, retention and deletion
Server location: [[להשלמה: מיקום השרתים]]
| Type of information | How long it is kept |
|---|---|
| Account details | For as long as the account is active. After the account is deleted: [[להשלמה: תוך כמה זמן נמחקים נתוני חשבון שנמחק]] |
| Runs, findings, letters and calendar events | For as long as the account is active, or until you delete them. After the account is deleted: [[להשלמה: תוך כמה זמן נמחקים נתוני חשבון שנמחק]] |
| Server logs and audit log (without content) | [[להשלמה: תקופת שמירת יומני שרת ויומן ביקורת]] |
| Encrypted backups | [[להשלמה: תקופת שמירת גיבויים]] |
| Free demo data | [[להשלמה: תקופת שמירת נתוני הדמו]] |
| Waitlist details | [[להשלמה: תקופת שמירת פרטי רשימת ההמתנה]] |
| Level 0 data on your computer | With you only. We have no access to it and we cannot delete it. |
Information that is deleted from the system may remain in backups until the end of the backup period.
9. Your rights
Under the Privacy Protection Law you have the right to:
- inspect the information held about you (section 13 of the Law);
- request correction of information that is not correct (section 14 of the Law);
- request deletion of information, in the cases the law provides;
- withdraw consent you gave, for example to notifications, to connecting Google Calendar or to receiving marketing messages.
Deleting an account. You can delete your account directly from the app. Deletion removes your data from the server. If you are the sole owner of the agency, all of the agency’s data on the server is deleted too: its other users, findings, letters, calendar events and agency profile. The confirmation in the app says this explicitly. Data that is stored only on your computer is not deleted.
How to make a request. Contact the privacy protection officer, at the details in the last section. We will ask to verify your identity before we provide information, and we will respond within the time set by law.
If you are a client of an agent. We will not be able to identify you in our systems, because we do not have names or Israeli ID numbers. Contact your agent, who is the database owner with respect to you.
You may also file a complaint with the Privacy Protection Authority.
10. Information security
The service’s server has not launched yet. The following list describes the security measures that have been set for it, under the Privacy Protection Regulations (Data Security), 5777-2017, and not an existing state. Each item will be verified before launch. The security level of the database: [[להשלמה: רמת האבטחה שנקבעה לפי תקנות אבטחת מידע (בסיסית, בינונית או גבוהה)]]
- Planned: all communication with the server will be encrypted (HTTPS).
- Planned: sign-in only through Google, Apple or Microsoft, with no passwords. Access tokens will be short-lived, and refresh tokens will be stored only as a hash.
- Planned: separation between agencies. Every action on the server will be limited to the data of the user’s own agency, and permissions will be set by role.
- Planned: a check of every piece of data that reaches the server, before it is stored and before it is sent to the AI.
- Planned: the logs will not record the content of requests or responses.
- Planned: the access key to the AI service will be kept on the server only.
- Planned: the database will not be accessible from the external network, storage will be encrypted, and backups will be encrypted and taken daily.
- In the desktop app: level 0 data is stored encrypted, with a key that is protected by the operating system.
- Planned: a database definitions document and a procedure for handling security incidents.
[[להשלמה: לאמת כל סעיף ברשימה מול מסמך האבטחה של השרת (server/SECURITY.md) לפני הפרסום]]
No system is completely immune. If a serious security incident occurs, we will act as the law requires, including reporting to the Privacy Protection Authority, and in cases where it is required we will notify you too.
11. Your responsibility as an agent or agency
The agent or agency is the database owner of their clients’ details, and is the party responsible for them. In practice:
- You are responsible for complying with the Privacy Protection Law with respect to your clients: making sure you are entitled to use the report data, giving clients the notices that are required and handling their requests for inspection and correction.
- Coding reduces the information that leaves your side, but it does not relieve you of responsibility. Coded data may be considered information about an identifiable person with respect to whoever holds the key, that is, you. That is why we protect it and handle it as personal information.
- You are responsible for the security of your computer and devices (updates, password, backup) and for access to your Google, Apple or Microsoft accounts. Do not give access to your account to anyone who is not authorized.
- Do not enter identifying details of clients into free-text fields.
If your agency needs a data processing agreement (DPA) with us, contact us. We recommend consulting a lawyer or your own privacy protection officer about the obligations that apply to you.
12. The free web demo
The demo (the addresses that begin with /k/) works differently from the apps:
- The reports you upload are read and processed in your browser only. They are not sent to our server and are not stored by us.
- To suggest which column is which, a description of the first rows is sent to the server. Only header rows are sent as they are (short rows, with no digits, where most cells are familiar column names). Every other row is masked: letters are replaced with a uniform letter, digits with #, and amounts by order of magnitude only. The description is passed to Anthropic in order to suggest the columns, and is not stored by us. The suggestion is shown for your approval and you can correct it.
- What we do store: the column structure of reports (column names and their mapping to fields), so that we do not need to recognize it again; the personal link details as they were given to us (name, agency, phone); and usage data (when you signed in, how many checks you ran, and how many rows and findings there were in each check), without amounts, policy numbers or names.
- The demo does not use cookies. The personal link is in the address itself.
13. The waitlist
When you join the waitlist on the site we collect your full name, your mobile phone number, your email address (if you filled it in), your type of business (independent agent, small agency or large agency), the number of insurers you receive reports from, and the page and language where you signed up. Providing the details is not required by law, but without them we cannot add you to the list.
- We use the details to contact you about the product.
- Marketing messages (updates and offers) are sent only if you explicitly ticked the consent box for them, in accordance with section 30A of the Communications Law (Telecommunications and Broadcasting), 5742-1982. Consent is not required, and not giving consent does not prevent you from joining.
- So that we can prove consent, we store the date and time when you approved the privacy policy and when you approved marketing messages (if you did), and the version of the privacy policy text you approved (its date).
- Signing up again with the same phone number does not replace the details already stored, and can only withdraw consent to mailings, not add it. This way it is not possible to change another person’s details or consent.
- Every person on the list has a personal link to unsubscribe from mailings, with a random code that identifies them. Every marketing message will include this link, and unsubscribing turns off the consent to mailings. The link looks like this:
https://commika.com/en/unsubscribe?token=…, where what follows the equals sign is a random 64-character code. Opening the link only shows a button, and the unsubscribing happens when it is pressed, so that a preview of a message does not unsubscribe anyone. Whoever holds the link can unsubscribe that one person from mailings and do nothing else with it: the page shows no personal details. - We also store an encrypted code derived from your IP address, as it reaches us from the server that receives the requests (in the X-Real-IP header), and not the address itself. It cannot be converted back into the address. It is used only to limit the number of sign-ups from the same source, and the records of the attempts are deleted after a day.
- Retention period: [[להשלמה: תקופת שמירת פרטי רשימת ההמתנה]]
- You can ask to delete the details at any time from the privacy protection officer in the last section. Deletion is carried out by the service operator on the admin page, and is recorded in the activity log, with an ID only and without a name.
15. Changes to the policy
If we change the policy materially, we will notify you in advance in the app or by email, and update the date at the top of the page. The binding version is the one on the site.
16. Contact, privacy protection officer and contacting the Authority
- Operating entity: [[להשלמה: שם הישות המשפטית]], number [[להשלמה: מספר חברה או עוסק מורשה]]
- Address: [[להשלמה: כתובת]]
- Email: [[להשלמה: דוא״ל ליצירת קשר]]
- Phone: [[להשלמה: טלפון]]
Privacy protection officer
- Name: [[להשלמה: שם ממונה הגנת הפרטיות]]
- Email: [[להשלמה: דוא״ל ממונה הגנת הפרטיות]]
- Phone: [[להשלמה: טלפון ממונה הגנת הפרטיות]]
Database registration
[[להשלמה: האם נדרש רישום מאגר מידע לפי תיקון 13]]
You may also contact the Privacy Protection Authority at the Ministry of Justice.