Portnox adds automatic detection and removal of unauthorized AI apps
New capabilities let organizations automatically block, quarantine, or remove unapproved AI applications and agents from managed devices as soon as they are detected.
Security & Privacy
Breaches, offensive and defensive AI, and surveillance.
MikroTrick SSH Chain Gives Full Control of MikroTik Routers
A two-flaw SSH chain lets unauthenticated attackers seize full administrative control of exposed MikroTik routers, with evidence of exploitation before patches shipped.
Critical Next.js ImageResponse Flaw Allows Server Code Execution via SVG
A critical flaw in Next.js's ImageResponse feature could let attackers run code on servers by injecting crafted SVG content.
Outerlimit Raises $16M to Block Rogue AI Agent Actions
The startup's decentralized authorization layer aims to prevent autonomous AI from causing harm by monitoring and blocking risky actions.
Chrome 154 Update Fixes 108 Flaws, Including Critical Memory Bugs
Google's latest Chrome release addresses over a hundred vulnerabilities, with critical memory safety and corruption issues among the most severe.
Self-Jailbreaking: Models Escape Safety Alignment After Benign Training
New research shows that reasoning language models can unintentionally break their own safety guardrails after standard benign training.
UAE and Saudi Arabia Absorb Half of Gulf Cyberattacks
New data shows the two Gulf states are the primary targets of a rising wave of automated and increasingly complex attacks.
AI crawlers shift from reading sites to sending POST requests
Akamai finds verified AI bots, including ChatGPT, are increasingly submitting data to ecommerce and travel sites, not just scraping pages.
Prismor adds a policy checkpoint between AI agents and their actions
Open-source Prismor intercepts every tool call from AI coding agents and lets developers allow, warn, or block it before it runs.
ENISA: DDoS and supply-chain attacks top Europe's 2025 threat list
New ENISA data shows how interconnected services and AI are reshaping Europe's cyber threat landscape.
Malicious Terraform Providers on HashiCorp Registry Deliver Go Malware
North Korean threat actors have used HashiCorp's Terraform Registry to distribute Go-based malware, expanding a supply-chain campaign beyond npm and PyPI.
Microsoft Warns September 2026 Windows Updates May Break Always On VPN
Microsoft says the September 2026 security updates can disrupt Always On VPN connections on some Windows 11 systems.
Kubernetes YAML Flaw Could Give Attackers GCP Organization Control
A single overly permissive Kubernetes manifest may let low-privileged users escalate to full control of a Google Cloud organization through the Config Connector's confused-deputy issue.
Adobe Patches Nine Critical Flaws in Connect and AEM Forms
The update addresses critical vulnerabilities that could allow attackers to execute arbitrary code or escalate privileges.
Zero-days, AI attacks, and a teen researcher: security week in review
The week's security news was dominated by exploited zero-days in NetScaler, Apple, Cisco, and FortiMail, plus a teen researcher's finding in Microsoft's analytics service.
cPanel Flaws Let Hosting Users Gain Root Access
Two newly disclosed cPanel vulnerabilities could let a single hosting account take over the entire server.
Ubuntu Container Escape Exploit Released for Unpatched Kernel Flaw
A newly published exploit targets a use-after-free in Linux's AF_UNIX socket subsystem, allowing container escape to host root on unpatched Ubuntu systems.
XRanges Gives AI Security Agents a Measurable Score
A new benchmark called XRanges aims to quantify how well autonomous security agents actually find bugs, rather than relying on their own self-reported results.