The project has three key milestones: the alpha launch, which provides an API for the Llama2-70b model; the beta launch, which adds full implementation of TPM-based attestation; and the 1.0 launch, which will be a fully secure version ready for audit. Users can contribute to the project by forking it on GitHub and creating a new pull request, or by spreading the word on social media. The project is open to feedback and suggestions, and users can get in touch through various channels, including opening an issue on GitHub, joining the Discord community, or setting up a one-on-one meeting with a team member.
Key takeaways:
- BlindLlama is a project developed by Mithril Security that aims to make AI confidential and transparent by using Zero-trust AI APIs.
- The platform ensures users' data is never exposed to the developers and provides cryptographic proof that these controls are in place.
- BlindLlama is designed to help developers working with sensitive data to easily get started with LLMs by using managed AI APIs that abstract the hardware and software complexity of model deployment.
- BlindLlama is currently under development and does not yet have the full security features. The project has three key milestones planned: Alpha launch, Beta launch, and 1.0 launch.