Anthropic

Anthropic was founded in 2021 by siblings Dario Amodei and Daniela Amodei and a group of fellow OpenAI researchers who left over differences about how frontier AI should be built. Structured as a public benefit corporation, it develops the Claude family of models and has become OpenAI’s chief rival — backed by billions from Google and Amazon while publishing the safety research (constitutional AI, interpretability, responsible scaling) that defines the “build it carefully” school.

The both-hands position

Anthropic occupies the debate’s strangest ground: its CEO writes essays predicting AI could compress a century of scientific progress into a decade, while also warning it may erase half of entry-level white-collar jobs; the company lobbies against some AI regulation as miscalibrated while endorsing California’s frontier-transparency law. It settled the authors’ class action over training data for a record $1.5 billion — the largest copyright settlement in history — and was sued by Universal Music Group over lyrics, even as its models became the coding tools of choice for much of the software industry.

To the pause movement, Anthropic is proof the race cannot police itself — a safety company that keeps scaling. To the accelerationists, it is proof safety talk is compatible with shipping. Its own answer is that someone at the frontier has to demonstrate what responsible development looks like, because the frontier will be occupied either way.