Independent tests find DeepSeek AI model safer, but still vulnerable

China-based AI security firm Fangcun Yuedong said its report found DeepSeek’s latest AI model performed better than earlier versions in independent testing but still had vulnerabilities.