Kalam vs the Rest — Your Data Stays Yours
The AI assistants you use daily are quietly learning from your conversations. Here's what the big players do with your data — and the three hard commitments Kalam makes instead.
Why data privacy matters for AI
Every message you send to an AI assistant is a window into your thinking. You ask it to draft sensitive emails, reason through personal dilemmas, help with medical questions, and discuss confidential business matters. In the Arab world especially, conversations often carry cultural weight — family, religion, politics, professional reputation.
The question isn't just "is this conversation encrypted?" — it's "who reads it afterward, and what do they do with it?"
What big AI companies do with your data
OpenAI (ChatGPT)
By default, conversations are used to train future models unless you opt out in settings. OpenAI's privacy policy allows human reviewers to access conversations for safety and model improvement purposes. Enterprise customers get stronger protections, but free and Plus users are the product by default.
Google Gemini
Google's terms allow human reviewers to read conversations. Gemini data may be used to improve Google's AI products and services. Opting out requires navigating account-level settings that most users never see.
Meta AI
Meta AI is trained on data from Facebook and Instagram — including your social history, interests, and connections. Your conversations with Meta AI may feed back into the same training pipeline that powers their advertising engine.
What Kalam does differently
Three hard commitments
We never train on your conversations. Ever.
Your messages are used solely to generate your response. They are not stored for model training, not fed into future versions of Kalam, and not used for any analytics beyond basic operational metrics (message count, error rates).
We never share your data with third parties — even if asked, we refuse.
No data brokers. No advertising partners. No government requests fulfilled without a court order in the applicable jurisdiction. We do not monetize your conversations, directly or indirectly.
End-to-end encryption in transit (TLS 1.3) and at rest.
All traffic between your device and our servers is encrypted with TLS 1.3 — the same standard used by global banks. Messages stored in our database are encrypted at rest, meaning even a raw database breach would not expose plaintext conversations.
Why this matters for the Arab world
The General Data Protection Regulation (GDPR) covers EU residents. California has CCPA. But most of the Arab world operates in a regulatory gap: no comprehensive data protection law, no independent privacy authority with real enforcement power, and no meaningful recourse when a foreign tech company misuses your data.
Arab users are not a priority for these companies' privacy teams. Arabic-language conversations are less likely to be reviewed by native speakers, meaning cultural context and sensitive personal information may be processed by reviewers who don't understand the weight of what they're reading.
Professional conversations in Arabic often involve personal relationships, family business dynamics, religious considerations, and social codes that have no equivalent in Western corporate culture. These conversations deserve privacy protection that was designed with them in mind — not bolted on as an afterthought.
Our promise
Kalam exists because the Arab world deserves AI that was built for it — not sold to it. Privacy is not a feature. It is the foundation. Every architectural decision we make starts with the same question: would we be comfortable if the user knew exactly what we were doing with their data?
The answer is always yes. Because the answer is always: nothing.
Start a private conversation →