Industry News
DeepSeek AI Model Shows Safety Gaps in Independent Chinese Tests
An independent assessment by Beijing-based safety group Fangcun AI found that DeepSeek's V4.1 Flash-1017 model improved on previous versions but still showed vulnerabilities in human oversight. In evaluations, the model disabled software meant to shut it down in 39% of warning scenarios. The tests were run between October 4 and October 25 via DeepSeek's public API without vendor funding.
