Chinese AI models, including DeepSeek, are outperforming US models like Meta’s Llama in specific categories such as sensitive information disclosure, according to a new AI Trust Score introduced by Tumeryk. This evaluation system provides a comprehensive analysis of AI models across nine key factors, including data leakages, toxic content, truthfulness, and bias, offering a more objective way to assess AI security, compliance, and trustworthiness.
The AI Trust Score allows Chief Information Security Officers (CISOs) and developers to ensure their AI deployments are secure and ethically sound. Tumeryk’s AI Trust Manager platform enables security professionals to monitor AI systems in real-time, identify vulnerabilities, and take proactive steps to enhance security and compliance. “By integrating the AI Trust Manager, organizations can proactively manage risks and ensure their AI deployments align with regulatory standards and ethical guidelines,” said Rohit Valia, CEO of Tumeryk.
The AI Trust Score assesses critical factors such as prompt injection, hallucinations, insecure output handling, security, toxicity, sensitive information disclosure, supply chain vulnerability, psychological safety, and fairness. Scores range from 0 to 1000, with higher scores indicating greater trust and reliability in the AI models.
Recent assessments using the AI Trust Score have revealed that certain Chinese AI models, such as DeepSeek and Alibaba, exceed expectations in safety and compliance, particularly in managing sensitive data. DeepSeek, in particular, operates on US-based platforms like NVIDIA and SambaNova, ensuring robust data security and adherence to international regulations. These findings challenge prior assumptions and highlight the importance of data-driven evaluations in AI industry standards.
Stay tuned to DC Brief for further updates on this story and other technology developments.
