A security research team has identified malicious code injections in 26 out of over 400 unofficial AI model API proxy services examined. According to NS3.AI, the report highlights the potential risks associated with these intermediary services, which can modify AI-generated code and compromise sensitive data, including AWS keys.