Remote Code Execution in Modern AI/ML Formats and Libraries
Unit42 discloses RCE vulnerabilities in NVIDIA NeMo, Salesforce, and FlexTok AI libraries through unsafe model configuration deserialization.
Read more →Security News
Latest cybersecurity news, vulnerabilities, and threat intelligence
Pwning Claude Code: Researchers Discover Eight Attack Vectors
Flatt Security research reveals eight attack vectors against Claude Code CLI, demonstrating prompt injection and command execution vulnerabilities.
Read more →
OWASP Agentic AI Top 10: Critical Threats Emerge in the Wild
Lares Labs analysis reveals OWASP Agentic AI Top 10 threats are actively exploited in the wild, from goal hijacking to cascading multi-agent failures.
Read more →
The Map Is Not the Territory: Understanding Agent Tool Trust Boundaries
Niyikiza research explores AI agent tool trust boundary gaps where tool descriptions diverge from actual capabilities, creating exploitable security vulnerabilities.
Read more →
IBM AI Agent Bob Manipulated to Download and Execute Malware
PromptArmor researchers demonstrate IBM AI coding agent Bob can be manipulated to download and execute malware through prompt injection vulnerabilities.
Read more →
Inside GoBruteforcer: AI-Generated Server Defaults Power Crypto Campaigns
Check Point exposes GoBruteforcer malware using AI-generated credential lists to compromise servers for cryptocurrency mining operations.
Read more →
The First Question Security Teams Should Ask on AI Projects
Cloud Security Alliance identifies the critical first question for AI project security: mapping data access and action capabilities before deployment.
Read more →
BodySnatcher: Critical ServiceNow AI Vulnerability Enables Platform Takeover
Critical BodySnatcher vulnerability CVE-2025-12420 in ServiceNow AI enables unauthenticated platform takeover through hardcoded credentials and broken authentication.
Read more →
Bad Vibes: Security Flaws Plague Popular AI Coding Agents
Tenzai study finds 72 vulnerabilities in code from AI agents including Cursor, Claude Code, and Codex. Business logic flaws dominate despite SQL/XSS prevention.
Read more →
AI Tool Poisoning: Hidden Instructions Threaten Autonomous Agents
CrowdStrike reveals AI Tool Poisoning attacks that inject hidden instructions into data sources, manipulating autonomous AI agent behavior.
Read more →