Anthropic's Bold Bet on Safety Overtakes OpenAI

Summary

A dramatic shift in the AI landscape has seen Anthropic, founded by former OpenAI researchers, strategically challenge and even surpass its former employer. Driven by a fear of 'scaling loss' – the exponential intelligence gains from increased model size, data, and compute – Dario Amodei and colleagues left OpenAI in 2020. While OpenAI focused on rapid consumer adoption, exemplified by ChatGPT's 900 million weekly users, Anthropic prioritized safety and enterprise applications with its AI, Claude. Despite starting with significantly fewer users, Anthropic’s revenue growth accelerated dramatically. By April 2026, Anthropic reached a $47 billion annual run rate, surpassing OpenAI's $25 billion, even though OpenAI had 30 times more users. This success is attributed to Anthropic's unique approach, including establishing itself as a public benefit corporation with a long-term benefit trust, prioritizing mission over profit. Crucially, Anthropic developed 'constitutional AI,' a self-grading system that uses a rulebook instead of human feedback (RLHF), which is expensive and prone to bias. This allows Claude to scale infinitely and prioritize truthfulness. Consequently, Anthropic secured substantial backing and focused on enterprise clients, with its average business customer spending 15x more than OpenAI’s. This rivalry, balancing power with safety, is pushing AI development forward.