Former OpenAI Researcher Issues Grim Warning, Says Only 5,000 People Have Access to Frontier AI While 2 B

Former OpenAI researcher Andrej Karpathy warned that the gap between public understanding of artificial intelligence and its most advanced capabilities is widening, with a small group of professionals witnessing a rapid leap in what AI agents can accomplish.

Billions Have Limited Exposure to Advanced AI

On Sunday, in a post on X, Karpathy said the gap in understanding AI capabilities is "widening," with billions of people barely exposed to large language models (LLMs) while others use advanced systems to complete complex tasks.

He estimated that around six billion people have barely interacted with LLMs, while another one to two billion use free-tier AI tools casually, often treating them as "a better Google search" or a writing aid.

By contrast, Karpathy estimated that roughly 20 million professionals, particularly in mathematics and software engineering, use frontier-grade models to complete projects that once took weeks or months.

AI Agents Take on Massive Projects

Karpathy said an even smaller group of about 5,000 people has internal access to the most advanced systems, which can coordinate thousands of AI agents on massive software projects, cybersecurity operations and mathematical research.