Security
Security and data privacy at Aidfolk
Your content is used to answer your customers, and nothing else. Here is exactly how we protect it, what we share with AI providers, and what we keep.
Each workspace's data kept apart
- Your knowledge passages are separated by the database itself (PostgreSQL row-level security), and the app connects as a database user that cannot bypass it.
- Everything else you create is scoped to your workspace by the application: other workspaces can't reach it.
Encryption
- All traffic uses HTTPS, and browsers are told to use nothing else.
- AI provider keys and the credentials of API sources are encrypted before they're stored, and never shown again after you save them.
- Passwords are hashed with Argon2, a modern algorithm designed to resist cracking.
Private file storage
- Uploaded files live in a private Cloudflare R2 bucket, never on a public address.
- Downloads use signed links that expire after a few minutes.
Account protection
- Email addresses are verified, and throwaway inboxes are refused.
- A bot check (Cloudflare Turnstile) guards sign-up and password reset, and sign-in attempts are rate limited.
- Sessions use secure, HttpOnly cookies. You can see every device you're signed in on and sign the others out; changing your password does it for you.
A chat widget that can't be borrowed
- Each agent answers only on the websites you list (subdomains included); anywhere else it refuses.
- Visitors get short-lived signed tokens tied to the agent, the conversation and the website, and requests are rate limited per visitor and per agent.
- The widget sets no cookies on your visitors, and it displays answers as text, with only web and email links allowed.
What AI providers receive
- To answer a question, the AI model receives the question, the passages from your content that match it, your agent's instructions and the last few turns of the conversation. Nothing else from your workspace.
- To index your content, its text is sent to OpenAI's embeddings API. Answers come from the model the agent uses: OpenAI, Anthropic or Google.
- On Pro, answers can run on your own AI provider account, under your agreement with that provider.
Who at Aidfolk can see your data
- Our admin tools hide conversation text by default. Reading a conversation takes a deliberate action with an optional reason, and every read is recorded with who and when.
- Visitors' IP addresses are never shown in those tools.
What we keep, and for how long
- IP addresses stored with sign-ins and widget chats are erased after 90 days. We keep only the approximate place (country, region, city), looked up on our own server: addresses never leave it for that.
- What you delete (conversations, knowledge, team members you remove) leaves the app at once and is erased for good from the database 30 days later. Server logs record only the network part of an IP address, never the full address.
- The database is backed up every night to private storage.
- To have your data erased completely, write to privacy@aidfolk.com.
Compliance, honestly
- Aidfolk doesn't hold a SOC 2 or ISO 27001 certification yet. If your company needs a security questionnaire answered, email us and we'll answer it.
- Payments are handled by Lemon Squeezy, our merchant of record: we never see or store card details.
Found a security problem?
Please tell us before anyone else: email support@aidfolk.com with “Security” in the subject. We'll confirm we've received it and keep you updated. The details of how we handle personal data are in our privacy policy.
Give your customers answers in seconds
Start on the Free plan with 50 AI credits a month. No card needed, and your agent can be live on your site today.