September 2024 Summaries

Clarifai has introduced several new features and improvements aimed at enhancing the platform's capabilities in concept management, AI model protection, and user interface. A significant update is the ability to define hierarchical and semantic links between concepts, which aids in building knowledge graphs and improving computer vision and natural language processing. The Prompt-Guard-86M model has been launched to protect LLM-powered applications from prompt attacks such as injections and jailbreaks by classifying inputs into benign, injection, and jailbreak categories. Other updates include the ability to download original assets from the Input-Viewer, enhanced annotation identification, and the deprecation of certain agent system operators. Additionally, improvements have been made to secure data hosting, labeling task management, and user interface adjustments for a better experience on smaller screens, including automatic resizing of bounding boxes and polygons to fit within canvas edges.