About

Why SecureMind exists

SecureMind started as a personal project, built out of curiosity and a desire to have more control over how conversations with AI models are handled.

Generative AI tools like ChatGPT have become incredibly useful, but there were aspects of the experience I wanted to explore further. How could sensitive information be better protected before it's shared? Why use the same model for every question when simpler requests could be handled by smaller, less expensive models? And how could we have more visibility into usage, costs, and what happens behind the scenes?

Those questions became the foundation for SecureMind.

A different approach to AI

The idea was fairly simple: create an AI workspace that brings together conversations, documents, and memory, while providing more control over privacy, model selection, and costs.

Privacy was one of the biggest considerations. It is easy to accidentally include personal or sensitive information in a message, especially when copying content from documents or other sources.

SecureMind checks chat messages and attachments for personally identifiable information (PII) and other sensitive data before anything is sent to an AI provider.

• Chat messages are checked in the browser. When something is detected, you see what would be replaced and decide before sending. The original values that are detected and replaced are not sent on, saved or logged. Only the cleaned message goes to the model.
• Attached files are checked when they are added, after the text is extracted and before it is saved, indexed or sent to a model. The text SecureMind keeps and uses, including the pieces used for AI processing, has the detected values replaced.
• Two things to know about files. On a Docker install the uploaded file itself is also kept on your computer, exactly as you uploaded it, so you can download it again. And an image has to be read by a model that can see pictures, so the image itself goes to that provider before its text is checked.

Another goal was to make model use more flexible. Not every question needs the most powerful or expensive model. SecureMind works with several AI providers and can choose a model by how demanding a request is, while showing which models were used and their estimated costs.

The goal was not to reinvent AI, but to build a workspace that offers more visibility into how it is being used.

From a local experiment to the cloud

The project began as something meant to run locally with Docker. It was a chance to experiment with different models, build out the features, and understand how the pieces work together.

As the application developed, taking it to the cloud seemed like a natural next step. A big part of that challenge was designing an AWS setup that could stay available on the web with very little ongoing infrastructure cost. With a serverless approach, the aim was to keep idle costs minimal, so that most of the spending comes from AI model requests.

It became as much an exercise in cloud architecture, security, and cost optimization as it was an AI project.

Open source and built to be shared

SecureMind is an open-source project for anyone who wants a more organized and transparent way to work with AI.

Whether you are a developer experimenting with different models, a technical professional working with documents and data, or simply someone who wants more control over AI conversations, SecureMind offers a workspace to explore those possibilities.

It is free to use and can be run locally or deployed in your own cloud account. The source code is public for anyone who wants to understand how it works, contribute improvements, or adapt it to their own needs. The idea is to make SecureMind something others can use, explore, and help shape.

A note on security

Security is a central consideration in SecureMind, but detecting sensitive data relies on algorithms and pattern matching that cannot guarantee every instance will be found.

Redaction is done on a best-effort basis to help reduce the risk of accidentally sharing sensitive information. SecureMind is not a compliance or regulatory certification product, and it does not guarantee complete protection against data exposure.

You are responsible for reviewing your information and deciding whether SecureMind is suitable for your intended use.