The Critical Stack

CISA has confirmed that attackers are exploiting a WordPress core flaw during software upgrades, with federal agencies due to patch by November 6, so check your WordPress estate now; meanwhile Node-RED is being exploited in the wild and AI-assisted intrusion tooling keeps showing up in real incidents.

Operational technology and AI tooling are both under active attack today: patch exploited Siemens IoT2050 and Langflow exposures, vet your IDE extensions, and assume the supermarket cold-chain attacks will repeat.

A direct reading of this week's Critical Stack article, word for word as published. Read the original at: https://thecriticalstack.com

Attackers are chaining edge-device and CI/CD flaws while AI agents now sit on both sides of the fight: ransomware crews and infostealers use them to scale, and your own agents need hard permission boundaries and tighter key custody.

Credentials and trust are the soft spot today: a CVSS 10.0 LXD/MicroCloud flaw lets attackers pose as trusted clients, and the Northflank breach shows stolen AI API keys are being abused, so patch your container infrastructure and rotate secrets now.

Edge and developer infrastructure (Ivanti, GitLab, Linux MPTCP) is being exploited or weaponized within days of patches, so patch internet-facing and CI/CD systems first and treat AI chat tooling as part of your attack surface.

Binance traces a roughly $6.3 billion crypto theft to a single compromised DevOps laptop, and the rest of today's briefing points the same way: developer tooling, AI model repositories and the software supply chain are where attackers are getting in, so treat engineer endpoints and extensions as tier-zero assets.

Attackers are going after the network and virtualization layer (SD-WAN, ESX, firewall management planes) to take encryption keys and blind your logs, so take management interfaces off the internet and assume edge devices can be compromised.

Enterprise edge and business-app stacks (Oracle, Adobe Commerce, Node) are being mass-exploited while AI-driven offense and agent-related exposure keep rising, so patch internet-facing systems and audit what your AI agents and vendors can reach.

Exploitation windows keep shrinking (PHP2Shell is already under attack, and exploitation averages three days after disclosure), so patch internet-facing PHP, edge and identity components first, and treat AI sandbox and reasoning-boundary failures as live risks.