Trust is part of the product.
To run your operations, you give us access to things you give no one else. So every piece of that access is encrypted, limited to the people who need it, and logged.
- AES-256 encryption for every password
- On-device verification before every view
- Two-step sign-in for every staff login
- Every action is logged
- Your data never trains models
What you give us, and what happens to it.
Your passwords don't travel over WhatsApp.
Credentials for your systems live in a dedicated vault. Not in messages, not in documents, not in personal password managers.
- Each record is encrypted on its own with AES-256-GCM, under a key that is not stored in the database and not included in backups.
- Only people assigned to your account can see that a record exists.
- Every password view requires identity verification on the Operations Manager's own device: fingerprint, face or Windows Hello.
- The password is shown for 30 seconds, then disappears.
- Every view is logged with who, when and for which task. Blocked attempts are logged too.
- You can ask for the list of views of your credentials at any time.
- A request to delete credentials is handled within one business day. When the engagement ends, they are deleted within three business days.
You see only your own. Technically, too.
Your portal identifies your account from the user's assignment on the server, not from anything that can be changed in a browser. One user cannot reach another client's data, even on purpose. Data is encrypted in transit and at rest, and the client area is closed to search engines. You can turn on two-step sign-in for your own portal access too.
We take notes, so you never repeat yourself.
Working calls are recorded and transcribed. That is how every commitment enters the system, and how a change of Operations Manager doesn't reset the context. Transcripts are kept with business-tier vendors only, and are never used to train models.
The AI remembers. It doesn't decide, and it doesn't see everything.
Our AI systems summarize calls, track commitments and flag what is about to slip. Decisions stay with people.
What the AI sees
- Transcripts of working calls
- Tasks, updates and notes
- Documents your Operations Manager passes to it as part of a task
What the AI never sees
- Usernames and passwords for your systems. It has no technical path to them, and an automated check stops any change that would open one.
- Tasks and materials you asked us to keep away from AI.
No training.
We work with Anthropic and OpenAI on paid plans only, and every account is set so your data is never used to train models.
A person approves.
Anything that leaves the system, a message, a payment or a booking, is approved by your Operations Manager.
You decide.
Ask, and a task, a document or everything of yours is marked "no AI" in our system. Nothing marked goes into our AI-assisted work, and your Operations Manager handles it personally.
Few people, each named, each with limited access.
A personal account for everyone.
No shared users. Every action is tied to the person who took it.
Two-step sign-in, every time.
Everyone at Ankora signs in with a password and a confirmation on their own device: fingerprint or face. A password alone does not open the system.
Access by assignment.
An Operations Manager sees only the clients assigned to them. Any other client simply doesn't exist for them.
Confidentiality in writing.
Everyone who touches your data has signed a personal confidentiality undertaking.
When someone leaves, their access doesn't stay.
Their permissions are revoked in our system at once, and every open session is closed. Access to external tools is closed within three business days. And you receive from us a list of every credential of yours they viewed in the last 90 days, so you can rotate them.
Under the hood.
Encryption
TLS on every access path, with HSTS. Encryption at rest in the database. User passwords are stored as bcrypt hashes, and reset links are single-use and expire.
Audit log
Every material action is logged with who, when, the state before and after, IP address and browser. Time entries keep a full version history.
Client isolation
The client's identity is derived on the server from the user's assignment, on every request.
Hardening
A dedicated content security policy for the client area with no third-party source at all. Rate limiting on sign-in and on every sensitive entry point.
Backup
A full backup every night. The restore script is tested automatically on every code change, and a full restore drill runs every quarter. Backups are deleted after 35 days.
Reviews
A monthly security review against the OWASP Top 10, and continuous tracking of vulnerable dependencies. Every finding is recorded with a severity and a fix date.
Built to work with your legal counsel.
When you pass personal data to us, the legal responsibility stays with you. Our Data Processing Addendum is written so you can meet it. We work to:
- Israel's Protection of Privacy Law, including Amendment 13
- The Protection of Privacy (Data Security) Regulations, 2017
- GDPR Article 28, for clients with activity in the EU
- OWASP Top 10, in development and in reviews
Every vendor, by name.
The application and the database are hosted in the United States. Email and documents are kept on Google Workspace's global infrastructure. Messaging channels are connected through Unipile, which stores data in France and routes encrypted traffic through intermediate servers in the European Union, Israel, the United States and Singapore. The phone line and call recordings go through Aloha, on the Peri Telecom network, and are stored in Israel. You will be notified before any new vendor is added.
| Vendor | Role | Location | Privacy policy |
|---|---|---|---|
| Vercel | Website and application hosting | United States | View policy Vercel |
| Neon | Database | United States | View policy Neon |
| Upstash | Rate limiting | United States | View policy Upstash |
| Resend | Email delivery | United States | View policy Resend |
| Google Workspace | Email, documents and backups | Global | View policy Google Workspace |
| Fireflies | Call transcription | United States | View policy Fireflies |
| Unipile | Messaging channel connection | France, with intermediate servers | View policy Unipile |
| Aloha | Phone line and call recording, on the Peri Telecom network | Israel | View policy Aloha |
| Anthropic | AI (Claude) | United States | View policy Anthropic |
| OpenAI | AI and recording transcription | United States | View policy OpenAI |
| Google Analytics | Marketing site measurement | Global | View policy Google Analytics |
You hear from us within 24 hours.
If we discover a security incident that touches your data, we notify you within 24 hours of learning about it, without waiting for the investigation to finish. You get what happened, which data may be involved, what we have already done, and what we recommend you do. We also report to the Privacy Protection Authority where the law requires it.
When we last checked.
Last security review: October 2026
Page last updated: October 2026
Changelog
- October 2026Personal confidentiality undertaking, signed by the whole team
- October 2026Two-step sign-in, required for all staff and available to clients
- October 2026On-device verification before every password view
- October 2026Encrypted credentials vault, with every view logged
- October 2026Full nightly backup with an automated restore test
- September 2026Audit log with IP address and browser for every action
- September 2026Data Processing Addendum published
Everything is written down, and available.
What people usually ask us.
Have a question we didn't answer?
We'll go through it on the intro call, together with whoever handles security on your side.