JSON data extracted from the public comments to the AI Action Plan RFI (Jun 2)

We used large language models to classify each comment along two dimensions: institution type and AI worldview. Google’s Gemini 1.5 Flash was accessed via the API interface and prompted to parse the RFI responses to extract the following information: • Submitter name and type • Institution type (big tech, small tech, civil society) • Specific organizational tag (startup, nonprofit, government, media, etc.) • Two AI worldview labels (from a set of six possible options) • Justifications for the institutional and AI worldview classifications • A filter tag indicating whether the comment should be retained or discarded