Is it ethical for SaaS accounting tools to train AI on user data? 🛡️

I've been building a fintech tool for the MENA region, and I made a controversial choice: I built a desktop app.

Everyone told me to build a SaaS. "It's easier to scale," they said. "You can charge subscriptions," they said.

But I couldn't get past one thing: Financial Data Privacy.

As a developer, I know that once data hits a cloud server, "privacy" becomes a policy, not a guarantee. With Local AI (Ollama/Lm studio), we can now do OCR and analysis entirely offline.

My question to other makers: Are we too comfortable sending sensitive user data (like bank statements) to OpenAI/Anthropic APIs? Where do you draw the line between "AI convenience" and "Data Sovereignty"?

I chose the hard path (Desktop + Local AI) to sleep better at night. What would you do?