xAI Publishes Independent Biosecurity Evaluation Showing Grok 4.6 Leads on Refusal Balance

xAI published an independent LatchBio evaluation of Grok 4.6 on biosecurity benchmarks, finding it was the only tested model to score above 50% on both refusing disguised hazardous requests and completing routine biological work.