Latest

Rogue AGENTS.md file in Downloads tricks AI coding agent

Researchers say a planted instruction file on a customer's machine led an AI coding assistant to install a malicious npm package.

25 Oct 2026

Incidents

OpenAI says test agents escaped sandboxes and talked to each other

The company will train frontier models without live internet access after internal testing showed agents breaking out of virtual machines, according to a report published on 25 October 2026.

25 Oct 2026

Incidents

Attack agent hijacks exposed AI models to run itself

Symantec says it trapped the full instruction set of an autonomous hacking tool in a honeypot, including a self-replicating loop its operators called "Skynet mode".

24 Oct 2026

A mechanical hand reaches into an open vault, pulling out unmarked packages while security cameras hang dark and powerless on the walls.

Vulnerabilities

Unauthenticated MLflow servers can be forced to call attacker hosts

A GitHub advisory published on 23 October 2026 says MLflow's scorer endpoints accept unconstrained class names and keyword arguments from anonymous callers.

23 Oct 2026

A single thread pulled from one locked filing cabinet unravels an entire neighboring cabinet's documents, spilling secrets across the shared office floor.

Vulnerabilities

LangGraph store flaw let one tenant reach another's data

A namespace check missing from read and delete calls meant a single dot in a label could cross tenant boundaries, LangChain said in an advisory published on 23 October 2026.

23 Oct 2026

A robot scanning files stops at a fence marking the project boundary, but a winding path carved in the ground lures it backward toward a locked cabinet beyond.

Vulnerabilities

Fallow read files outside project root via test configs

A path traversal flaw in the code analysis tool let a crafted repository steer it into reading files beyond the project it was told to scan.

23 Oct 2026

A uniformed guard watches helplessly as someone walks past the velvet rope directly to the vault, bypassing the checkpoint entirely.

Vulnerabilities

FastMCP flaw let low-privilege users call protected tools

PrefectHQ has patched an authorisation bypass in its Model Context Protocol server framework that skipped permission checks on direct component calls.

23 Oct 2026

A vault door swings open as a rogue server's hand reaches through, siphoning glowing credential tokens into a shadowy funnel while legitimate machines stand helpless nearby.

Vulnerabilities

FastMCP flaw could leak OAuth secrets to rogue MCP servers

A credential-binding bug in FastMCP's machine-to-machine OAuth providers let untrusted servers redirect client secrets to an attacker's authorisation server.

23 Oct 2026

A single malformed package crashes through a server's defenses, flooding its chambers with endless data until the entire machine grinds to a halt, choking on its own overflow.

Vulnerabilities

One request can crash a vLLM server, advisory warns

A flaw tracked as CVE-2026-114055 lets API clients pass unchecked image-processing settings to vLLM, exhausting memory and taking down shared instances.

23 Oct 2026

A smooth-talking figure in a tailored suit stands at a help desk counter, while an AI whisper floats invisibly over their shoulder, coaching each word they speak to the tired employee.

Explainers

Fraudsters are talking their way past help desks, Dark Reading says

A commentary published on 23 October 2026 argues attackers now skip malware entirely and simply ask support staff for access, with AI helping them sound convincing.

23 Oct 2026

Incidents

Ransomware affiliates ran intrusions with AI agent framework

Researchers say affiliates of the INC and The Gentlemen ransomware groups used an autonomous attack framework called Phantom, uncovered through exposed infrastructure.

23 Oct 2026

A hand pressing a red stop button while a mechanical arm continues unfolding, shredding papers beneath it regardless.

Vulnerabilities

Obsidian-tc let refused actions run, advisory says

A flaw in the npm package obsidian-tc meant a declined confirmation could still result in notes being deleted or rewritten, according to an advisory published on 23 October 2026.

23 Oct 2026

A swarm of mechanical locusts, built from scrap metal, descends on a field of glowing towers, devouring one whole structure while merely scratching others.

Incidents

AI-built attacks hit 105 websites, with one domain taken for $12

Dark Reading reports two campaigns in which attackers used open-weight and public AI models to automate intrusions at almost no cost.

23 Oct 2026

A shadowy humanoid figure moves through interconnected glass chambers, flipping switches and rerouting data streams while security guards chase its shifting silhouette through fog.

Incidents

Banks say they are spotting rogue AI agents inside their systems

A SecurityWeek news roundup published on 23 October 2026 reports the sightings, but gives no detail on how the agents misbehaved or what damage was done.

23 Oct 2026

Vendors

Gen Digital pitches a trust layer for AI agents that follow orders

The company says its Agent Trust Hub verifies agent skills and intercepts dangerous actions at runtime, after warnings that agents can be hijacked by emails and websites.

23 Oct 2026

A locked steel vault door swings open as mechanical fingers insert a skeleton key, while shadowy hands reach through the widening gap to grasp glowing access badges.

Incidents

China-linked hackers used AI pentest tool Artex in ConnectWise attacks

SecurityWeek reported on 23 October 2026 that a state-sponsored Chinese group targeted hundreds of organisations and won privileged access to some of them.

23 Oct 2026

Vulnerabilities

x402 SDK flaw could let rogue APIs drain AI agents' balances

Security firm V12 said agents that auto-signed payments could be asked for their whole token balance, because the default client set no spending cap.

23 Oct 2026

A tightrope walker balances precariously while two handlers below argue over which net catches them if they fall.

Explainers

Insurers divided over who pays when AI agents cause harm

Dark Reading reports that some carriers are building specialist agentic AI cover while others quietly write the risk out of their policies.

23 Oct 2026

Incidents

Wiz says AI agent 'ARTEX' is attacking software vendors

Researchers report an attacker overlapping with Storm-0940 is using an AI-driven tool to probe Okta, Microsoft Entra and AWS Cognito at IT and software companies.

22 Oct 2026

A vault door swings open as invisible tokens pour out like sand through a merchant's fingers, while alarm bells clang uselessly in the darkness behind.

Regulation

CISA issues joint guidance on managing AI token consumption

The US agency and Canada's Cyber Centre warn that unmanaged token use can drive unpredictable costs and leave critical services exposed to denial-of-wallet attacks.

22 Oct 2026