← Back to App
Privacy Policy
Last updated: September 27, 2026
Talk to Data is a product of LogicGridz LLC ("Talk to Data," "we," "us"). This policy describes how we handle data for the Talk to Data multi-tenant SaaS platform.
1. What We Collect
| Data Type | Purpose | Retention |
|---|
| Email, name, password (hashed) | Account authentication | Until account deletion |
| Chat messages & queries | Conversation context, session history | Until account deletion |
| AI agent learnings & corrections | Improve accuracy for your organization | Until account deletion |
| Uploaded data (CSV/Excel) | Enable querying your data | Until you delete the dataset |
| Telemetry (query count, response time) | Performance monitoring, usage analytics | Indefinite (anonymized) |
| IP address | Rate limiting, security | 30 days |
| Payment info | Billing (processed by Stripe) | Managed by Stripe |
2. What We Do NOT Collect
- We do NOT persist result rows from your connected external databases to disk — queries run live against your database, and results are held only briefly in server memory (a short-lived cache) before being discarded.
- We do NOT use your data to train AI models
- We do NOT sell or share your data with third parties for marketing
- We do NOT track you across other websites
3. Third-Party Services
| Service | Purpose | Data Shared |
|---|
| Stripe | Payment processing | Email, payment method (not stored by us) |
| Anthropic (Claude), OpenAI (GPT-5), Google (Gemini), DeepSeek | AI query processing (via YOUR API keys — BYOAK) | Your questions + dataset schema and representative column values (see Section 4) |
| Neon (PostgreSQL) | Application database | All app data (encrypted at rest) |
| Render | Server hosting | Server logs (no customer data) |
| Sentry | Error tracking | Error stack traces, URL, HTTP method, and user ID — not request/query content |
| Resend | Email notifications | Email addresses for alerts and transactional email |
| Plausible | Privacy-friendly, cookieless website analytics | Aggregate visit data only (page views, referrer, country-level location) — no cookies, no personal identifiers, no customer/query data |
| AWS, Snowflake, Databricks, ClickHouse, Microsoft SQL Server, MySQL (as applicable) | Only if YOU configure a connector to one of these as an external data source | Connection credentials (encrypted at rest — see Section 5) and live query traffic to the data source you configure. We do not use these services ourselves unless you connect to them. |
4. AI & Your Data
- When you ask a question, we send the AI provider more than a bare schema: column names and data types, plus representative values so the AI can map your wording to the right column — for text/categorical columns this includes actual distinct values (up to ~40 per column) present in your data, and for numeric columns a min/max range. Your question itself is also sent.
- Your full dataset (all rows) is NOT sent to AI providers — the sample/representative values described above are a subset, not a full export, but they are real values from your data, not synthetic placeholders. If your data contains sensitive values in low-cardinality text columns (e.g., a status or category field), be aware those values may appear in the AI provider's request.
- AI providers process your requests under their own privacy policies. We use "Bring Your Own API Key" so you have a direct relationship with the AI provider — we do not see or store your provider bill, and your API key is encrypted at rest on our side (see Section 5).
- Agent learnings (corrections) are stored in our database and injected into future prompts for accuracy.
5. Data Security
- Passwords hashed with bcrypt
- Optional two-factor authentication (2FA) using an authenticator app (TOTP standard — compatible with Google Authenticator, Authy, 1Password, and similar apps). You can enable this from Settings. We never see or store the codes your app generates — verification happens locally by comparing against your account's own encrypted secret.
- Database connector credentials and BYOAK AI provider keys encrypted at rest (AES-256-GCM, an authenticated cipher), including Databricks access tokens.
- All traffic encrypted via HTTPS (TLS 1.2+)
- JWT authentication with short-lived access tokens (15 min) + refresh tokens (7 days)
- Rate limiting at multiple levels (IP, user, organization, AI concurrency), scaled to your plan tier
- Multi-tenant data isolation enforced at the application layer by organization ID on every query and API route, verified by an automated test suite on every release. This is application-level tenant isolation, not database-native row-level security (e.g., Postgres RLS policies) — we use the term "row filtering" for policies you configure yourself (see below) to avoid confusion with that database feature.
- Optional customer-configured row filters and column masking (e.g., restrict a role to only see rows matching their region, or hide a column entirely) — separate from the tenant isolation above, which is not configurable and always on.
- Immutable audit log for administrative actions
6. Your Rights
- Access: You can view your data via the app (chat history, agents, datasets). Admins can also export usage/telemetry data (users, token usage, agent learnings, datasets) as CSV from the Governance dashboard.
- Deletion (dataset-level): You can delete individual datasets or agents yourself, immediately, via the app — this removes the underlying stored data right away.
- Deletion (account/organization-level): Full account/organization deletion is currently a manual process — email [email protected]. We do not yet offer a fully self-service "delete my account" button. We will delete your organization's data within 30 days of a verified request.
- Portability: Your data (chat history, agent configs, datasets) can be exported on request by emailing [email protected]; broader self-service export beyond admin CSV exports is on our roadmap but not yet available.
7. Cookies & Analytics
We use only essential localStorage for authentication (JWT token storage) — no advertising cookies, no cross-site tracking cookies. For website analytics we use Plausible, a privacy-friendly, cookieless analytics tool that measures aggregate traffic (page views, referral sources, country-level location) without cookies, without collecting personal data, and without tracking you across other websites. Because it stores no personal identifiers and sets no cookies, no cookie-consent banner is required.
8. Children
The Service is not intended for anyone under 18. We do not knowingly collect data from minors.
9. Data Processing Agreement
If your organization is subject to GDPR, CCPA, or similar data protection law and needs a formal Data Processing Agreement covering how we process personal data on your behalf, see our Data Processing Agreement or email [email protected] to request one.
10. Changes
We may update this policy. Changes will be posted here with an updated date. Material changes will be communicated via email.
11. Contact
Privacy questions: [email protected]. Talk to Data is a product of LogicGridz LLC.