Every day, people copy names, IBANs, contracts and medical reports into AI tools. The providers see everything, because all content ends up on their servers. AUSTR.AI is designed to prevent that, or at least limit it. Three classic use cases that many people know and have probably used in exactly this form:
"Summarise this contract"
A sales representative copies a customer contract into ChatGPT, with names, IBANs and amounts. All of it now sits on OpenAI's servers.
"Translate this medical report"
A doctor has an AI translate a medical report. Patient name, social security number, diagnosis: everything goes to the provider, and the provider decides what is stored there and for how long.
"Help me with my tax return"
A freelancer copies income data, tax number and address into the chat. Google, OpenAI or Anthropic now have this data.
How AUSTR.AI works
The idea is simple: AUSTR.AI is a local chat assistant that checks each of your inputs for sensitive data on your own computer (or server). Detection runs locally. A preview shows you what will be anonymised. You can correct it by hand and only send the message once all sensitive data has been replaced. The content then goes, anonymised, to the language model of your choice: OpenAI, Claude, Gemini, Mistral or a local model via Ollama or LM Studio. Only the anonymised data is processed there. When the answer comes back, AUSTR.AI puts your real data back in, on your computer. The software handles the rest automatically.
Like ChatGPT, but your sensitive data stays with you
AUSTR.AI runs independently and locally on your own instance, meaning your computer or your server. The mapping between code names and real data never leaves your device. No message goes out without your click on Send, and if anonymisation fails, nothing is sent.
Why AUSTR.AI was built
The tool was not developed to be marketed, but to give smaller companies and private users more control over their data.