Security Digest — 2026-03-07
AI is reshaping both sides of cybersecurity today: defenders are deploying models to find vulnerabilities at scale, while threat actors are leveraging AI to accelerate and broaden every stage of their attacks.
AI Security Research
OpenAI Codex Security Scanned 1.2 Million Commits and Found 10,561 High-Severity Issues The Hacker News
OpenAI launched Codex Security, an AI-powered agent that finds, validates, and proposes fixes for vulnerabilities in codebases. In a research preview for Pro, Enterprise, Business, and Edu users, it scanned 1.2 million commits and surfaced over 10,000 high-severity issues.
Microsoft: Hackers Abusing AI at Every Stage of Cyberattacks BleepingComputer
Microsoft reports that threat actors are increasingly integrating AI across the full cyberattack lifecycle — accelerating reconnaissance, scaling phishing, and lowering technical barriers for less-skilled attackers. The trend signals a structural shift in offensive capabilities.
Anthropic Finds 22 Firefox Vulnerabilities Using Claude Opus 4.6 AI Model The Hacker News
In a two-week security research partnership with Mozilla, Anthropic used Claude Opus 4.6 to discover 22 previously unknown Firefox vulnerabilities — 14 rated high severity. All issues were patched in Firefox 148, demonstrating AI-assisted fuzzing and code analysis at production scale.
Vulnerabilities & Exploits
Termite Ransomware Breaches Linked to ClickFix CastleRAT Attacks BleepingComputer
The ransomware group tracked as Velvet Tempest is combining the ClickFix social-engineering technique with legitimate Windows utilities to drop DonutLoader malware and the CastleRAT backdoor, linking a wave of recent breaches to coordinated Termite ransomware campaigns.