Independent tests find DeepSeek model security improved, but flaws remain

A report by Chinese AI security company Fangcun Yuedong says DeepSeek’s latest model performed better than earlier versions in independent security tests, but still has vulnerabilities.