Machine Intelligence Research Institute

Founded in 2000 by Eliezer Yudkowsky as the Singularity Institute and renamed in 2013, the Berkeley-based Machine Intelligence Research Institute (MIRI) invented much of the vocabulary of AI alignment — decision theory, agent foundations, the “friendly AI” problem — decades before frontier labs existed. For most of its history it was a pure research shop, funded by donors including Peter Thiel and Vitalik Buterin.

From alignment research to “shut it down”

In January 2024, CEO Malo Bourgon announced a strategic reversal: having concluded its technical agenda could not solve alignment before frontier capabilities arrived, MIRI declared that “policy and communications will be a higher priority for MIRI than research going forward,” aiming to push the world’s major governments toward an international agreement halting progress toward smarter-than-human AI.

The pivot’s flagship arrived in September 2025: “If Anyone Builds It, Everyone Dies,” by Yudkowsky and MIRI president Nate Soares, became a New York Times bestseller. Its thesis is maximal — that building superintelligence with anything like current techniques kills everyone, and that governments should treat frontier AI hardware like fissile material, monitoring even small clusters of high-end GPUs. Reviewers were split between calling it the decade’s most important warning and criticizing its certainty; most working AI researchers assign far lower probabilities to the doom scenario, and critics note MIRI’s own two-decade research program produced few results by its own admission.

MIRI today is less a lab than a lobby for stopping the technology its founders once hoped to perfect — the most absolutist position in the AI debate, held by its oldest institution.