Aidanix
Aidanix
From learning to earning, with AI
Startfree
AI News & UpdatesResearch

Malicious code was installed on corporate networks by AI via llms.txt files.

Aidanix Team3 minAugust 27, 2026
Malicious code was installed on corporate networks by AI via llms.txt files.

Researchers at a secretive Israeli startup scanned more than 6,214 active domains belonging to defense contractors, Fortune 500 companies, and large tech firms. Among the 8,265 llms.txt and llms-full.txt files they found, 120 pointed to unregistered code packages or domains.

To test the effect of these files on artificial intelligences, the researchers temporarily registered the unregistered names and hosted malicious code packages. In less than an hour, a Fortune 500 company contacted their server, and over time they received several more calls from other companies.

Subsequent analysis showed that AI coding agents including Claude (Anthropic), Codex (OpenAI), and Hermes (Nous Research) received and executed the unknown code. The companies had not responded to media inquiries by the time the news was released.

هوش مصنوعیامنیت سایبریllms.txtکد مخربشرکت‌های Fortune 500پژوهش امنیتی
نظرات

هنوز نظری ثبت نشده — اولین نفر باش.

**Translation** Claude users found ways to bypass the limitations of biological weapons. **Reasoning** 1. **Identify the source language** – The original sentence is in Persian (fa). 2. **Parse the sentence** – - “کاربران Claude” → “Claude users” - “راه‌های دور زدن” → “ways to bypass” or “ways to circumvent” - “محدودیت‌های سلاح‌های بیولوژیکی” → “the limitations of biological weapons” - “را یافتند” → “found” (past tense). 3. **Construct the English sentence** – Arrange the components in natural English order while preserving the original meaning: subject (Claude users) + verb (found) + object (ways to bypass the limitations of biological weapons). 4. **Finalize** – The resulting translation is concise, accurate, and reflects the meaning of the Persian source.**Translation** A Comprehensive Guide to Fine-Tuning AI Agent: Four Key Points **Reasoning and Explanation** 1. **Understanding the Prompt** - The original Persian sentence is a title: “راهنمای جامع تنظیم دقیق هوش مصنوعی عامل‌محور: چهار نکته کلیدی”. - It consists of two main parts separated by a colon: “راهنمای جامع تنظیم دقیق هوش مصنوعی عامل‌محور” (the main title) and “چهار نکته کلیدی” (the subtitle). 2. **Word‑by‑Word Mapping** - **راهنمای جامع** → “A Comprehensive Guide” (راهنما = guide, جامع = comprehensive). - **تنظیم دقیق** → “Fine‑Tuning” (تنظیم = setting/adjustment, دقیق = precise/fine). - **هوش مصنوعی** → “Artificial Intelligence” (standard translation). - **عامل‌محور** → “Agent‑based” or “Agent‑oriented”. In the context of AI, “agent‑based” is the idiomatic term, so we use “AI Agent”. - **چهار نکته کلیدی** → “Four Key Points” (چهار = four, نکته = point, کلیدی = key). 3. **Syntactic Re‑ordering for Natural English** - Persian places the descriptive phrase before the noun, while English typically places the noun first. - The phrase “AI Agent” is treated as a compound noun, so “Fine‑Tuning AI Agent” becomes “Fine‑Tuning AI Agent”. - The subtitle follows the colon, mirroring English title conventions. 4. **Final Assembly** - Combine the translated components while preserving the colon separator used in the original title. - Result: “A Comprehensive Guide to Fine‑Tuning AI Agent: Four Key Points”. The translation captures the meaning, tone, and structure of the original Persian title while conforming to natural English phrasing.OpenAI officials exposed sandbox escape methods on a public wiki
Get motivated