top of page

RELEASES

A chart on Craig A. Kaplan’s desk shows existential risk from SuperIntelligence falling from about 10% to 0.01% with safety by design, beside books by Simon, Newell, and Pinker.

October 21, 2026

Dr. Craig A. Kaplan sorts SI risk into three categories, explains why loss of control is the only one without precedent, and reviews where experts put the odds of catastrophe. Then he makes the case that aligning SI with human values and designing it to become safer as it gets smarter can bring existential risk down to the level of ordinary human violence: one in ten thousand.

Dark navy infographic titled “Slower is not safer. Smarter is safer.” showing four principles of Democratic SI: a green-and-blue network for “The safest SI is a network”; a glowing blue cube with a question mark for “Black boxes need accountability”; a gold shield and connected nodes for “Good SI agents must check bad actors”; and blue human figures with a heart symbol for “Humans supply the values.”

October 16, 2026

Dr. Craig A. Kaplan argues that slowing SuperIntelligence development is not necessarily safer. Drawing on expert risk estimates and recent AI advances, he explains why accelerating the development of safer systems is essential and outlines four design principles that could dramatically reduce existential risk as SuperIntelligence becomes more powerful.

bottom of page