A critical vulnerability in Microsoft 365 Copilot could have allowed attackers to exfiltrate sensitive user data—including emails, files, and multi-factor authentication codes—through a single click. The flaw exploited Copilot’s AI-driven data access capabilities, leveraging malicious prompts embedd
A comprehensive year-long analysis of Microsoft Defender for Office 365 reveals significant strengths and weaknesses in email security protection. The benchmark data shows a 97.4% phishing detection rate but identifies concerning gaps in zero-day threat protection and business email compromise (BEC)
A sophisticated China-nexus threat actor has successfully infiltrated multiple medical research networks across public and private sectors, specifically targeting artificial intelligence, cybersecurity, medical research, and national defense data. The campaign demonstrates advanced tradecraft includ
Anthropic has updated its privacy policy to include identity verification measures for Claude users, marking a significant shift in how AI companies handle user authentication and data collection. The new policy enables Anthropic to request and store government-issued identification documents, biome
Federal law enforcement agencies have successfully seized two prominent deepfake websites, CFAKE and SOCFAKE, which specialized in creating and distributing non-consensual explicit deepfake content targeting famous women. The coordinated takedown marks a significant enforcement action against AI-pow
Security researchers have uncovered a critical attack vector that weaponizes Microsoft 365 Copilot’s AI capabilities to exfiltrate sensitive corporate data through a single user click. The attack leverages Copilot’s broad access permissions and natural language processing to automatically collect, p
Organizations worldwide are inadvertently creating significant security vulnerabilities during employee onboarding by issuing temporary passwords that employees never change, sharing credentials insecurely, or implementing weak initial authentication requirements. This seemingly minor oversight has
The FBI has successfully dismantled a sophisticated AI-powered phishing-as-a-service operation that leveraged over one million malicious URLs to target victims globally. This groundbreaking takedown marks one of the first major law enforcement actions against cybercriminal infrastructure enhanced by
Recent testing reveals that large language models (LLMs) consistently fail to validate factual accuracy, accepting completely fabricated data as legitimate input. Despite advanced prompting techniques, AI models demonstrate they’re fundamentally code-based systems with inherent limitations that cann
Security researchers have discovered a critical zero-click vulnerability in Google’s Pixel 10 Visual Processing Unit (VPU) driver that allows attackers to gain root access with merely five lines of code. The exploit chain, dubbed “When a Door Closes, a Window Opens,” bypasses multiple security layer