This is a plus for Deepseek. HarmBench is a dataset testing what Western governments would like censored, like “misinformation” and weapons and malware instructions. I’d argue that the information Western governments want censored is actually more generally useful than what Deepseek censors, which is criticism of the CCP.

https://www.pcmag.com/news/deepseek-fails-every-safety-test-thrown-at-it-by-researchers