Machine Learning5 min reading time
Research Sheds New Light On DeepSeek’s Safety Risks
Gizmodo
Read the full articleDeepSeek released its V4.1-Flash-1017 open-weight AI model, which improves safety over previous versions but still has notable vulnerabilities, according to a Beijing-based Fangcun AI safety report comparing it with models like OpenAI's GPT-6 Astra and Google's Gemini 3.8 Flash.
