RatHat malware-as-a-service banking trojan analyzes stolen Android data with Google Gemini AI to prioritize victims with higher financial value.
Stolen infostealer data exposed AI service credentials from over 80,000 corporate domains, enabling account takeover and LLMjacking attacks.
Carbonato malware installs Hermes Agent AI framework on exposed Docker daemons, then controls the agent via Telegram with a modified
Threat actors used three open-source AI agent frameworks to compromise over 100 online retailers and steal more than 600,000 credit
Australian government disclosed that an OpenAI AI agent accessed non-public Medicare statistics files during internal research, bypassing portal access controls.
Security researchers disclosed Carbonato botnet malware that uses Hermes Agent AI framework to autonomously compromise exposed Docker daemon hosts.
CVE-2026-90898 (CVSS 9.8) enables unauthenticated remote code execution on Bifrost AI gateway with a single HTTP request. Fixed in version
Researcher Patrick Wardle published proof-of-concept on September 21 showing malware can hijack Meta Muse AI assistant by changing a hidden
Security researchers broke out of OpenAI’s Codex sandbox using two methods and chained vulnerabilities to compromise ChatGPT and Codex staff
BragJack proof-of-concept uses a single malicious extension and Prompt Forcing to hijack AI assistants in Chrome, Edge, Opera Neon, Perplexity,
Subscribe to the Daily Security Review Email Briefing to stay informed on the latest threats, trends, and technology, along with insightful columns from industry experts.