MIT Technology Review reports that researchers at Stanford and MIT have launched the AI Observatory, an independent platform analyzing real AI conversations, after finding that companies like Anthropic and OpenAI selectively publish usage data that favors their own products. An analysis of over 24,000 conversations found that work-related use makes up a smaller share of AI chatbot activity than company reports suggest, with conversations about health, relationships and other sensitive topics appearing far more often. The researchers also found substantial differences in how models are used in practice — for example, Grok for news and politics, Claude for coding and ChatGPT for homework — that published company reports do not capture.
