No Standard Defines What to Record for an LLM Call, So I Built One
Johannes built a recording system for LLM calls to enhance AI traceability and reliability.
Johannes is developing an investigations engine that requires every finding to trace back to its source. In a court case, an expert witness using an AI assistant failed to recall what he input or the sources used, prompting the court to question the reliability of AI outputs. This incident highlighted the need for proper logging of AI interactions. Johannes has since built a recording system to ensure that such issues do not arise in the future, providing crucial information about model calls.