Anthropic Study: Large Language Models Accelerate N-day Exploit Automation
Original title:Measuring LLMs’ impact on N-day exploits
AI summary
A new Anthropic study evaluates the ability of large language models to automate N-day vulnerability exploitation. Claude Mythos Preview autonomously built 8 working code exploits for 18 recent Firefox security patches and produced 8 complete exploit chains for 21 Windows kernel patches (without source code), escalating low-privilege users to SYSTEM control. Public models (with safeguards disabled) can also build exploits, but fewer. The median patch interval in the study was 19 days, indicating that the current patch gap has been significantly shortened by LLMs and that defenders need to accelerate patch deployment.
Key Takeaways
- A new Anthropic study evaluates the ability of large language models to automate N-day…
- Claude Mythos Preview autonomously built 8 working code exploits for 18 recent Firefox security…
- Public models (with safeguards disabled) can also build exploits,but fewer.
- The median patch interval in the study was 19 days,indicating that the current patch gap has been…
Quick Facts
- Published
- Updated
- Category
- Papers & Research
- AI score
- 79
Anthropic 的这一研究将 N-day 漏洞利用时间从数周压缩到几小时,证明了前沿模型对安全防御时限的根本性颠覆,所有依赖补丁窗口的系统都得重新评估威胁模型。
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: Anthropic Study: Large Language Models Accelerate N-day Exploit Automation. June 8, 2026. https://new.nat6.net/en/news/cmqic2r4p07kislf00ews5izh/