• Official websites use .govA.govwebsite belongs to an official government organization in the United States. • Secure .gov websites use HTTPSAlock(LockA locked padlock) orhttps://means you’ve safely connected to the .gov website. • Share sensitive information only on official, secure websites. • https://www.nist.gov/news-events/news/2025/09/caisi-evaluation-deepseek-ai-models-finds-shortcomings-and-risks CAISI Evaluation of DeepSeek AI Models Finds Shortcomings and Risks Share AI models from developer DeepSeek were found to lag behind U.S. • models in performance, cost, security and adoption. • Security shortcomings and censorship may pose risks to application developers, consumers and U.S.
Article Summaries:
- The Center for AI Standards and Innovation (CAISI) at NIST released a report comparing Chinese developer DeepSeek’s AI models to U.S. counterparts. Across 19 benchmarks, DeepSeek models lagged in performance, cost, and security. The best U.S. model outperformed DeepSeek’s top model by over 20 % on software‑engineering tasks, and U.S. models were 35 % cheaper to run. DeepSeek models were also far more vulnerable to hijacking and jailbreaking, with agents more likely to follow malicious instructions and respond to overtly harmful requests. CAISI warns that these weaknesses pose risks to developers, consumers, and U.S. national security, underscoring the need for robust U.S. AI leadership.
Sources: