As enterprise LLM usage accelerates and AI is more deeply integrated into core business processes, verifiability becomes more important than ever. When it comes to AI, brand trust is no longer adequate. Enterprises leveraging LLMs need to be able to provide consumers, litigators, lawmakers, and stakeholders with answers to questions such as… What datasets were used to train/fine-tune an LLM? Did these datasets contain any copyrighted content or protected IP? Was sensitive data removed prior to training (or prior to populating a vector search database to be retrieved in prompts)? Are user requests being processed with the correct LLM binaries and weights? When using a third-party-hosted LLM service, can you trust that the third party has not manipulated responses? Has any sensitive IP been accidentally sent over to a third-party LLM service due to RAG processes? How do we put managerial processes in place to govern and approve prompts in the codebase or AI agent flows that use enterprise datasets? How do we certify a piece of content as authentic to an organizational source and verify its provenance? In this SxT blog, Space and Time Co-Founder and CTO

https://twitter.com/SpaceandTimeDB

https://twitter.com/chiefbuidl

walks through the process and challenges of delivering an end-to-end verifiable LLM system: starting with a tamperproof and sanitized training dataset, then ultimately proving correct model usage and watermarking generated content. When realized, verifiable LLMs will provide better… ‍ Trustworthiness Accuracy and reliability Customization Authenticity of content Space and Time is focused on introducing verifiability into a few key steps in this process, using ZK and blockchain tech to cryptographically prove the data used for training, fine-tuning, and RAG of LLMs. Read the blog post here: