OpenAI Codex Security Scanned 1.2 Million Commits and Found 10,561 High-Severity Issues

A broad set of cybersecurity developments: AI is being used both to find flaws and to build malware—OpenAI’s new Codex Security scanned 1.2M commits and flagged 10,561 high-severity issues while threat actors (e.g., Transparent Tribe) and researchers (Anthropic using Claude Opus) used LLMs to discover or mass-produce vulnerabilities and implants. Multiple active campaigns and new malware families were reported (VOID#GEIST delivering XWorm/AsyncRAT/Xeno RAT; ClickFix campaign deploying Lumma Stealer; MuddyWater’s Dindoor backdoor; Dust Specter’s SPLITDROP/GHOSTFORM; APT28’s BadPaw/MeowMeow; APT