Reports published by artificial intelligence companies based on their own user data prevent independent researchers and policymakers from accessing transparent information, leading to decisions being made on incomplete data. Developed by researchers from Stanford and other academic institutions, a new independent platform called the "AI Observatory" analyzes over 25,000 real user chats, offering an unbiased framework of how popular AI tools like ChatGPT, Claude, and Grok are actually used.
Why Do Independent Data Differ from AI Companies' Reports?
Major AI developers such as Anthropic and OpenAI frequently publish economic indices and reports regarding the usage of their products; however, these reports largely focus on business and productivity-oriented scenarios. When the AI Observatory team applied Anthropic's filtering methods to their own independent dataset, they discovered that approximately 48% of the chats were excluded due to these filters.
An examination of these filtered areas revealed that non-business personal use encompassed sensitive topics such as health and relationships (44.2%), harassment and hate-related content (27.5%), and sexually explicit content (16.7%). Contrary to reports by major corporations, independent platforms demonstrate that the rate of AI usage in personal and sensitive areas is significantly higher.
User Profiles Vary by Model
The research revealed that the intended use of AI tools differs significantly from model to model. Accordingly:
- Grok: Users predominantly engage in interactions centered around news, politics, and disinformation.
- Gemini: Primarily preferred for roleplay scenarios.
- Claude: Stands out in software development and coding processes.
- ChatGPT: Serves as the go-to destination for homework and general consumer queries.
One-sided reports published by a single company fail to fully reflect this diversity and the differences in usage between models.
Sectoral Reflections and the Importance of Independent Data
For researchers examining the social and economic impacts of AI technologies, access to transparent and unbiased datasets is of critical importance for the reliability of future policy decisions. Platforms like the AI Observatory break the data monopoly shaped by corporate interests, paving the way for a healthier analysis of actual usage habits within the AI ecosystem.
Frequently Asked Questions
What data sources and methods does the AI Observatory project use?
The platform aggregates and analyzes approximately 25,000 real AI chats obtained from seven distinct existing datasets, collected with user consent.
Why are reports published by major AI companies criticized?
These reports are criticized because they typically focus only on business- and productivity-oriented use cases while filtering out personal, sensitive, or potentially rule-breaking interactions, thereby presenting an incomplete and one-sided picture.
*This news article was prepared based on data published by MIT Tech Review — AI.
💬 Comments
No comments yet. Be the first!
You must be logged in to comment.
🔑 Log In