#ai-security

OpenAI Launches GPT-5.4-Cyber for Defensive Security

OpenAI unveiled GPT-5.4-Cyber, a specialized AI model optimized for defensive cybersecurity operations and threat detection capabilities.

Anthropic Launches Project Glasswing AI Security Initiative

Anthropic unveils Project Glasswing using Claude Mythos AI model to detect vulnerabilities with AWS, Apple, and security firms.

AI Systems Vulnerable to Hidden Web-Based Command Injection

Researchers discover AI models can be tricked into executing malicious instructions hidden on attacker-controlled web pages, potentially exposing sensitive data.

Font Rendering Attack Bypasses AI Assistant Security Filters

Researchers discover new font-rendering attack that hides malicious commands from AI assistants using specially crafted HTML fonts.

OpenAI Launches Codex Security AI Agent for Vulnerability Detection

OpenAI released Codex Security, an AI-powered security agent that automatically finds and fixes code vulnerabilities for enterprise customers.

OpenClaw AI Critical RCE Flaw Patched — All Developers Must Update Immediately

OpenClaw developers released an emergency security patch on March 2, 2026, fixing a critical remote code execution (RCE) vulnerability in their widely-adopted AI agent development platform. The flaw, discovered via responsible disclosure, exposed AI systems to remote attack. Immediate update required.