Writing

AI alignment across borders

As Chinese open-source models surge in usage and benchmarks, how do they behave in potentially dangerous or sensitive scenarios?