We are on a mission to democratize confidential and trustable AI through enclave-based solutions. Enclaves are secure computing environments that combine confidentiality and verifiability to protect critical data during analysis. Learn more about them here!
BlindChat is a confidential & verifiable Conversational AI.
It is available via an easy-to-use web API similar to ChatGPT, Bard, or Claude, but with BlindChat, users have cryptographic guarantees that their prompts remain private from the AI provider (in this case Mithril Security). Not even our admins can access user data.
We achieve this by leveraging a privacy-enhacing technology called enclaves with our underlying core solution, BlindLlama.
Before each connection with BlindChat, we verify the system, providing technical proofs that:
You can test BlindChat here.
BlindLlama is the core technology behind BlindChat. It enables AI providers to deploy easy-to-use AI APIs within a privacy-by-design system using enclaves. With BlindLlama, AI providers can provide cryptograhic guarantees that they will not be able to access user data.
BlindChat Local is a variant of the BlindChat project, but instead of protecting user data by deploying our conversational AI server within an enclave - we eliminate the need for a remote server altogether! Instead, BlindChat Local operates entirely in your browser, using transformers.js for local inference and saving conversations in the browser cache - this means your data never leaves your device!
You can try BlindChat local here
Note that the models and performance with BlindChat Local are restricted by the user's device. For best performance, we recommend remote inference with our standard BlindChat solution.
Client-side bandwidth requirements | Client-side computing requirements | Model capabilities | Privacy | |
---|---|---|---|---|
β BlindChat with BlindLlama | Low | Low | High | High |
BlindChat Local | High | High | Low | High |
Regular AI APIs | Low | Low | High | β οΈ Low |
We would love to hear your feedback or suggestions, here are the ways you can reach us:
Want to hear more about our work on privacy in the field AI?
Thank you for your support!