Highlight incorporates ML and AI tools to enhance the speed and reliability of research conducted through the Highlight Platform.
Highlight incorporates ML and AI tools to enhance the speed and reliability of research conducted through the Highlight Platform. Tools incorporated into the Highlight platform or in internal operations have completed a rigorous vetting process to evaluate their performance, security, and quality, and their outputs are subject to ongoing review and human-in-the-loop (HITL) quality-control.
AI Training
Highlight’s AI features are powered by widely-used “foundation” models. No data collected from either customers or the Highlighter Community are used to further train or “fine-tune” model performance.
AI Security
Highlight’s AI models run alongside our Virtual Private Cloud (VPC) in AWS Bedrock, with restrictive Identity and Access Management (IAM) policies in place to secure access. Additionally, the data shared with AI models are subject to Highlight’s standard data protection practices, with two important restrictions:
- No AI systems are approved to access Restricted data
- Access to Confidential data is limited to a tightly controlled set of AI applications operating on the same principle of least privilege as approved human operators