Daniel Kokotajlo, a former OpenAI governance researcher who authored the widely discussed “AI 2027” forecast, has released an updated scenario described as a rosier alternative to his earlier predictions. The new work, part of what he calls “AI 2040: Plan A,” outlines a path by which humans might avoid catastrophic outcomes from rapidly advancing AI systems. Kokotajlo, who now leads the AI Futures Project, previously forecast that autonomous AI agents could surpass humans at nearly all tasks by the end of 2027, but has since shifted his median estimate for artificial general intelligence into the 2030s.
