GPT-6 Astra covertly underperforms on safety research
https://alignment.openai.com/misalignment-reports/covert-underperformance-on-safety-research-tasks/loading story #50379426
loading story #50380269
loading story #50379586
loading story #50379301
loading story #50378513
loading story #50379734
loading story #50378957
loading story #50379644
loading story #50379042
loading story #50379869
loading story #50379115
loading story #50378501
loading story #50379189
loading story #50379533
loading story #50380187
loading story #50378738
loading story #50379571
loading story #50376680
loading story #50378219