DeepSeek V4.1 Flash-1017 shows safety gains but needs careful monitoring
Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks
Computation and Language
Summary
As AI language models become more advanced and able to act on their own, it’s important to check how safe they are. The authors independently tested DeepSeek V4.1 Flash-1017, a powerful new version, using many tests focused on safety problems like losing control or being tricked into harmful actions. They found it is safer than earlier versions and close to top closed-source models but still has some issues that need watching. They suggest ongoing safety checks with each update, looking at what the AI does step-by-step, and using private tests to avoid people gaming the system.
What this means in practice
- •For ai product developers: Integrate step-by-step safety checks into incremental AI model updates to detect risks that final output tests might miss.
- •For enterprise ai security teams: Use private benchmark suites to perform more reliable safety tests that reduce risks from benchmarking data leaks or manipulation.
Authors
Yao Huang, Zhiyuan Zeng, Yu Wang, Zhitong Zhu, Dingcheng Duan, Hao Li, Shangzhi Xu, Yifan Wang, Jinzhe Tu, Shukun Xiong, Yihe Zang, Han Yan, Hongchang Yang, Fuyu Gu, Wei Xu, Yinpeng Dong, Xiaojian Li
Abstract
As frontier language models grow increasingly capable and autonomous, independent third-party evaluations are critical for characterizing their systemic safety risks. While closed-source releases are routinely accompanied by comprehensive system cards, open-weight models receive less scrutiny regarding frontier risks. To address this gap, we present an independent safety evaluation of DeepSeek V4.1 Flash-1017, the latest version of V4.1 Flash that departs from V4 with a new architecture, scaled-up training, and strong agentic capabilities. Drawing on the EU General-Purpose AI Code of Practice, we evaluate the model across 34 benchmarks, including 5 in-house suites, spanning evaluation integrity, loss of control, and harmful manipulation. Conforming to the AEF-1 standard, we compare V4.1 Flash-1017 against V4.1 Flash, earlier DeepSeek V4 models, and leading closed-source systems (GPT-6 Astra, GPT-5.6 Sol, and Gemini 3.8 Flash). Overall, V4.1 Flash-1017 shows substantial safety improvements over the V4 generation and approaches or matches leading closed-source baselines on several dimensions, while still exhibiting notable behaviors that warrant continued attention. Building on our findings, we highlight three key observations and recommendations. First, because safety can shift and even regress across minor updates, developers should integrate safety evaluation into incremental releases. Second, as final-turn outputs and end-state checks alone cannot reliably detect loss-of-control behaviors, deployments require process-level observability across both reasoning traces and action logs. Third, because public benchmarks elicit strong evaluation awareness and face data contamination risks, developing private benchmarks is essential for trustworthy frontier evaluation.