Security and privacy

Your conversations are your business. We treat them that way.

Chats hold prices, credit terms and personal messages. DataLense is built so you choose what it reads, where AI runs, and how long anything is kept, and every claim on this page is something the product does today.

The cautious setting is the default

AI is off for every new workspace. Every chat starts as Metrics only. Before anything reaches a model, phone numbers, emails, Telegram handles and links are masked and senders are pseudonymised.

Metrics only: Messages are stored and counted for reply times and volume. Nothing is sent to AI and no files are read.
Full analysis: Report files are read into records, and AI may use the chat to answer questions and suggest details.
Excluded: Nothing is kept. Existing messages from the chat are deleted and new ones are ignored.

How your data is kept

A private network: DataLense runs on AWS inside a private network. Provider events reach it only through verified, signed requests.
Encrypted storage: The database is encrypted at rest, backed up daily for 7 days and protected from deletion. File storage is private, with public access blocked.
Keys per customer: ERP credentials, webhook secrets and AI keys are encrypted with a separate data key for each customer account, wrapped by AWS KMS. Telegram logins are sealed with their own key that only the Telegram service can use.

Who can see what

Workspace isolation in the database: PostgreSQL row-level security keeps every workspace to its own data, and enforces each member's chat access, not just the app.
Roles and per-chat access: Owner, admin, member and viewer. Private sources, and “can reply” or “read only” per chat.
An audit trail: Every access change is logged. Invite links work once and expire after 7 days.
Sign-in: Google, a one-time email code that expires in 10 minutes, or passkeys. See and end your active sessions any time.

How long we keep it

Retention you set: Per workspace or per source: forever, 1 year, 6 months, 90 days or 30 days. Older messages are removed every night.
Real deletion: Exclude a chat and its messages are deleted. Remove a source and everything copied from it is deleted.
Signed webhooks: Incoming provider events are verified, and outgoing webhooks are signed with HMAC-SHA256 and guarded against requests to internal addresses.

Where AI runs is your call

Each workspace picks a data policy. DataLense never falls back to a location you haven't allowed.

DataLense servers only: An open model running on DataLense's own servers inside our private network. Nothing is sent to a model provider.
Enterprise cloud models: Claude models on Amazon Bedrock, called from within DataLense's own cloud account.
Any provider: Adds free models through OpenRouter for the lowest cost. Choose it for data you're comfortable sharing more widely.

Built on proven foundations

The infrastructure, standards and practices DataLense is built on. These describe how we build, not third-party certifications.

AWSInfrastructure in a private network, keys wrapped by AWS KMS
PostgreSQLRow-level security isolates every workspace
PasskeysPhishing-resistant sign-in to the FIDO standard
WebAuthnThe W3C standard behind passkey sign-in
Google sign-inOAuth sign-in with your Google account
OWASPWeb security practices guide how we build
GDPR and DPDPPrivacy laws we design data handling around

Straight answers

Can we get value without switching on AI?

Yes. The inbox, reply times, the list of customers waiting and one profile per person all work with AI off. Many teams start there and switch on Full analysis for the groups that carry orders and reports once they see the picture.

Who decides what DataLense reads?

Your workspace owners. Every chat has its own setting, every source can be made private to a few people, and every teammate sees only what you grant. Every change to access is logged.

Do you keep the questions we ask?

No. DataLense records how many questions were asked and what they cost, so you can track usage, but not the questions or the answers.

Can we take data out again?

Yes. Set a retention period per workspace or source, exclude any chat to delete what was kept from it, or remove a source to delete everything copied from it.

Try it with AI switched off

Link a number, keep every chat on Metrics only, and see your inbox and reply times first. Turn on more when you're comfortable.

Create your workspace