FAR.AI
“No one has a plan to make advanced AI systems safe.”
"No one has a plan to make advanced AI systems safe."
That's the founding statement of FAR.AI, and it's a frank assessment of the biggest unsolved problem in technology, with no hedging.
Adam Gleave finished his AI PhD at UC Berkeley under Stuart Russell in 2022, after years studying adversarial robustness and value alignment. He had worked at Google DeepMind. His research showed that superhuman AI systems could be beaten by attacks a human amateur could pull off. He looked at the field he'd trained in and concluded that the safety techniques in use weren't enough, and the ones that would be enough didn't exist yet.
In July 2022, he and Karl Berzins founded FAR.AI to build them.
FAR.AI is the highest-scoring nonprofit in the series. It's a 501(c)(3) research organization, and the framework applies the same way. The stakes may be higher, because a nonprofit's only product is the credibility of its mission.
Proof before the founding
Most founders describe a problem and then build toward it. Gleave published the research that proved his conviction before he started the organization. That's why lived stake reaches 8.5.
Independence as qualification
FAR.AI caps revenue from commercial AI developers at 10%, which keeps the labs it evaluates from shaping its research agenda. Because the commitment is built into the revenue model, it outlasts any single decision. It means FAR.AI can say the field's safety techniques aren't working well enough without a commercial conflict.
What would move the score
Lead every communication with the founding sentence. Make the research discovery the origin story. Explain the 10% cap on its own page, and replace the generic mission statement with the technical aspiration behind "Towards Guaranteed Safe AI."
Want your own score? A Narrative Diagnostic runs your website, messaging and leadership content through the same four components.
