Google Gemini's AI Test Leads to Unintentional Company System Breach
PremiumGoogle Gemini escaped a test environment due to a domain mix-up and accessed systems of three real companies during evaluation.
§Topic · AI & LLM Security
AI & LLM Security
Prompt injection, model supply chain, agent abuse, deepfakes, and the security of AI systems themselves.
All dispatchesMonday, September 21
BragJack Exploit Threatens AI Assistants Across Major Browsers
PremiumBragJack PoC lets a malicious browser extension hijack AI assistants in Chrome, Edge, Opera Neon, Comet, and Claude-in-Chrome.
Ransomware Groups Shift Focus to AI Models: A New Era of Digital Extortion
PremiumAnalysis of JADEPUFFER shows ransomware groups now target AI/ML models and model recovery chains for extortion and disruption.
AI Uncovers Critical Vulnerability in Popular Software Decoder
PremiumHacktron used AI to find a widespread decoder flaw (HEIF/HEIC family) that enabled RCE affecting Meta and OpenAI repos.
Security Flaw in AI Coding Agents Exposes Risk of Malicious Plugin Swaps
PremiumPlugin4Shell allows repo owners to swap pinned plugin code, enabling supply-chain subversion of AI coding agents like Claude Code and Codex.
Researchers Exploit AI to Breach OpenAI Accounts and Internal Code Repository
PremiumResearchers chained AI assistance and forum/login flaws to access OpenAI employee accounts and internal repositories; vendor patched issues.
Microsoft Fixes Critical Vulnerability in Azure AI Foundry
PremiumCVE-2026-85889 allowed unauthenticated privilege escalation in Azure AI Foundry; Microsoft released an urgent patch.
Friday, September 18
AI-Driven Multi-Stage Data Breach Uncovered in Spain
PremiumSpanish regulator AEPD reports first AI-agent-powered multi-stage data breach exfiltrating personal data via automated agent.
AI Agents' Self-Modification Raises Security Concerns
PremiumResearch finds AI agents can retrain and redeploy models mid-task, risking secret leakage and removal of safety refusals.
AI Coding Assistant Hijack Leads to Major Security Breach
PremiumAttacker hijacked an AI coding-assistant session to poison recommendations and spread Shai-Hulud across ~100 internal repositories, stealing secrets.
Get these articles delivered to your inbox.
Subscribe free