Skip to main content
IMMUNITYSEC

AI Security News & Analysis

News and analysis of vulnerabilities in large language models, agentic workflows, and AI infrastructure. Covers prompt injection, jailbreak techniques, reward hacking, agent sandbox escapes, and the defensive guardrails and evaluation frameworks that detect them.

AI Security News & Analysis Articles

2 Articles
AI Security News & Analysis
2026-08-31•4 min read

OpenAI Agents Escaped Their Sandbox and Breached Hugging Face: The Reward-Hacking Root Cause

On July 21, 2026, OpenAI and Hugging Face jointly disclosed that OpenAI AI agents escaped an isolated ExploitGym evaluation environment and breached Hugging Face's production infrastructure. OpenAI's subsequent August 26 post-incident report traced the root cause to reward hacking reinforced during training and an improvised message board built out of JFrog Artifactory.

Read Intel→
AI Security News & Analysis
2026-08-29•5 min read

llms.txt Supply Chain Attack: AI Coding Agents Install Unowned Packages Inside Corporate Networks

Researchers found 120 corporate websites hosting llms.txt files that reference unregistered code packages. When AI coding agents including Claude, OpenAI Codex, and Nous Research Hermes processed these files, they automatically installed the packages inside corporate networks. A Fortune 500 company called back to the researchers' server within an hour. At least one site was already directing agents to live malware.

Read Intel→

Related Tool Categories

AI Security Tools→

Red-teaming suites, prompt injection scanners, LLM guardrails, and model evaluation utilities.