Welcome to TrustWatch

TrustWatch is a community-driven platform where users share their experiences with AI models, along with a custom TrustScore that reflects their level of confidence in a model’s behavior. These reports are grouped into meaningful categories such as safety, fairness, and performance across diverse demographics and domains.

• Each report is weighted based on how recent it is and how many other users validate it, ensuring that the TrustScores remain dynamic and relevant.
• The TrustScore for a large language model (LLM) is calculated by aggregating multiple user-submitted reports over time. This continuous, longitudinal monitoring provides users and stakeholders with deep insight into how models perform in real-world conditions—highlighting both strengths and limitations before they are widely adopted.
• Our goal is to make TrustWatch a central hub for evaluating the safety and fairness of AI systems.