Breaches, vulnerabilities and the people cleaning up after them. Reporting for defenders, not for panic.
Mandiant and GreyNoise document active exploitation of CVE-2026-88771 and CVE-2026-88772 in NetScaler appliances, with threat actors deploying custom malware families and gaining root access across government and financial sectors.
The Dutch Institute for Vulnerability Disclosure fell victim to an autonomous AI-driven attack that exploited an unpatched flaw, leaving behind extensive forensic evidence of the agent's clumsy post-exploitation work.
Model Context Protocol makes it simple to wire AI agents into internal systems, but the real challenge lies in enforcing fine-grained permissions. GraphQL offers a proven way to define exactly what data and actions each agent can reach.
Researchers found that actual TLS configurations often diverge significantly from national security guidelines, prompting the development of TLSGatekeeper, a tool that enforces cryptographic standards at the network level without decrypting traffic.
OpenAI's research framework has uncovered a new class of prompt injection attacks that can propagate themselves through email, files, and code comments, instructing AI agents to reproduce the malicious payload across systems.
Keio Corporation, a major Japanese railway operator, fell victim to a ransomware attack over the weekend that knocked out business systems on its hospitality side. The incident prompted the company to shut down its network and file a police report.
Law enforcement in the Netherlands has confirmed the detention of a 24-year-old Amsterdam resident linked to the ShinyHunters hacking collective, with the suspect set to appear before a district court this week.
Security researchers discovered thousands of misconfigured Supabase instances exposing sensitive user data, passwords and tokens. The findings highlight growing risks from AI-assisted development tools.
