Karen Hao Says OpenAI, Anthropic Exploit Labor and Data as IPOs Loom in 2026-2027
Updated
Updated · Bloomberg · Jul 24
Karen Hao Says OpenAI, Anthropic Exploit Labor and Data as IPOs Loom in 2026-2027
3 articles · Updated · Bloomberg · Jul 24
Summary
Karen Hao said OpenAI and Anthropic could build AI with far less labor exploitation, data extraction and resource use, arguing today’s dominant model reflects choices about scale, profit and power rather than technical necessity.
25% of unemployed Americans now hold four-year degrees, she said, tying AI’s push to automate knowledge work to a shrinking market and a shift from full-time roles into poorly paid gig-style data labor.
Malta’s 1-year free ChatGPT Plus deal shows how governments can trade citizen data for access, Hao argued, saying frontier AI firms need fresh user data as high-quality internet training material runs short.
AlphaFold, used by more than 3 million people in 190 countries, was her example of a lower-cost 'bicycle' of AI—specialized, publicly grounded and socially useful without the same scale-driven harms.
With Anthropic potentially listing as soon as October and OpenAI leaning toward 2027, Hao said stronger governance must come from the public and elected officials, not founders or market forces alone.