A genealogy of AI safety: how directions are born, and how they die (2005-2026)

According to LessWrong, researcher Elena Ericheva has charted the evolution of AI safety from twenty-oh-five through mid-twenty twenty-six by documenting three hundred twenty-three events, one hundred twenty-nine key actors, and eighteen distinct research directions. The field's growth wasn't linear. Some directions merged into the mainstream; others faded. The center of gravity shifted from early philosophical concerns and isolated institutes toward empirical technical safety and governance research, concentrating around Anthropic and national safety institutes. A striking pattern: after twenty twenty-three, funding shifted from Open Philanthropy dominance to government support across multiple countries and venture capital—about two hundred sixty-eight million dollars into safety startups. Yet research attention and funding rarely align. Interpretability commanded roughly thirty-seven times more research attention than its funding would predict. The conclusion: AI safety didn't follow one path. It branched, consolidated, and drifted in ways a neat narrative cannot capture.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton