PromptArmor study finds open-weight models 88% more susceptible to indirect prompt injection
Original title:Open-weight models 88% more susceptible to cyber attacks
AI summary
In an evaluation of 9,800 trials, PromptArmor found that open-weight models (DeepSeek, Kimi, GLM, Qwen) were susceptible to real-world indirect prompt injection attack chains at a rate of 26.7%, versus 14.2% for closed-source models (GPT, Claude, Gemini), making open-weight models 88% more susceptible.
Key Takeaways
- In an evaluation of 9,800 trials,PromptArmor found that open-weight models…
- 7%,versus 14.
- 2% for closed-source models (GPT,Claude,Gemini),making open-weight models 88% more susceptible.
Quick Facts
- Published
- Updated
- Category
- Papers & Research
- AI score
- 67
原文基于 9800 次试验给出开放权重与闭源模型在间接提示词注入下的易感率差距和分攻击数据,方法与结果可直接用于安全选型参考。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: PromptArmor study finds open-weight models 88% more susceptible to indirect prompt injection. October 16, 2026. https://new.nat6.net/en/news/y1c13l53cd5u3wg565qu60q4d/