A new research project called the AI Observatory has compiled and analyzed real AI conversations with popular models like Claude and Gemini, providing an independent view of how people use generative AI. The Observatory's data, collected with users’ consent through seven existing datasets, aims to help researchers and policymakers better understand AI usage patterns. The project highlights that AI use varies significantly across models and has changed over time, with many more sensitive behaviors than are captured in reports from major AI companies. Source: mittr
The AI Observatory found that conversations filtered out by work-focused datasets like Anthropic’s Economic Index included more health, relationships, and adult topics. When the Observatory applied Anthropic’s methods to their dataset, they found that nearly half of the conversations—or 48%—would have been filtered out. These non-work-related conversations were more likely to include health and relationships (44.2% versus 31.2%), adult or illicit topics (7.9% versus 2.1%), harassment and hate (27.5% versus 5.66%), and sexual content (16.7% versus 2.4%). Source: mittr
The Observatory’s research also showed that AI use varied depending on the model. For example, Grok and Gemini were used more frequently for information retrieval, while Anthropic was used for coding, and Gemini for social and roleplay uses. The AI Observatory aggregated 24,521 conversations across 85,633 conversational turns from 5,000 users interacting with 52 different models between 2023 and 2025. Source: mittr
