We Still Don't Know How People Really Use AI — and It's Shaping Your Strategy Anyway

AI Summary

MIT Technology Review has highlighted a blind spot at the centre of the AI industry: everything we supposedly know about how people use ChatGPT and Claude comes from OpenAI and Anthropic themselves. AI researchers, including Stanford's Anka Reuel, point out that these companies only release the data they choose to, and — in her words — "there is no independent source to corroborate it." For Malaysian business leaders, the practical consequence is simple: the usage "trends" driving your AI strategy are vendor-curated marketing, not verified market research. The most reliable usage data available to you is the data you collect inside your own organisation.

Key Takeaways

• Anthropic and OpenAI regularly publish reports on how people use Claude and ChatGPT, but they control what gets released — researchers say there is no independent corroboration of any of it.
• These reports influence what businesses, investors, and governments believe AI is "actually used for," which means unverified data is quietly shaping procurement, hiring, and training decisions globally.
• For Malaysian firms, the gap is doubled: global reports say little about local usage patterns — Bahasa Malaysia prompts, manufacturing workflows, SME back-office tasks — so local leaders are extrapolating from a curated foreign sample.
• You cannot fix the industry's transparency problem, but you can fix your own: an internal usage audit gives you a dataset no vendor can spin.
• As AI shifts from chatbots to autonomous agents that take actions across systems, unverifiable usage claims get riskier — a wrong agent action executes, it doesn't just sit in a chat window.

What Happened

MIT Technology Review published a piece this week examining a question most of the industry skips past: do we actually know how people