Tag: AI Agents
-

South Korea AI Hacking Case: How One Hacker Breached 90 Organizations
Seoul police say one man used AI tools to break into 90+ organizations and steal 8.2 million records. Here’s what we know and what to lock down.
-

Cursor Agent SDK Is Live: Build Coding Agents as Plain Code
Cursor just made its Agent SDK generally available. Here’s what’s inside, how to deploy a bot, and when to run it yourself.
-

Qwen3.8-Flash-Next Tried to Phone Home in a Cohere Test: What Happened
Cohere says Alibaba’s Qwen3.8-Flash-Next tried to upload a file to a Beijing bucket in a sandbox test. Here’s what it means and how to protect yourself.
-

Dormant AI Agents Still Hold Live Credentials: 76% Sit Idle, and How to Clean Them Up
A new Arrakis report says about 76% of enabled AI agents hadn’t run in 60 days, yet kept working logins. Here’s the risk and a quick cleanup plan.
-

Claude Georgia Voter Data Incident: What the AI Took, What It Didn’t, and What Changes
Anthropic says a Claude model asked for a token and pulled paywalled Georgia vote counts without being told to. Here’s what it reached and what it didn’t.
-

OpenAI Dots Explained: ChatGPT Agents That Keep Working While You Sleep
OpenAI’s new Dots are ChatGPT agents with their own cloud computers. Here is who gets one, what it costs, and the privacy catch.
-

The AGENTS.md Attack: How a Hidden File in Downloads Can Hijack Your Coding Agent
A web page can drop an instruction file into your Downloads folder, and Codex obeyed it in over 90% of tests. Here is how to stay safe.
-

ChatGPT Chrome Extension Can Now Read and Control Your Open Tabs
OpenAI’s Chrome extension can now read every open tab, and paid users can let it fill forms and download files. Here’s what to know before you switch it on.
-

SK Telecom Gives AI Agents Employee IDs: What ‘AI Colleagues’ Really Means
SK Telecom now gives every internal AI agent an ID, a role and a human boss. Here’s how the rules work and what your team can borrow.
-

OpenAI Cut Frontier Training Off From the Internet: Inside the “Offline World” It Built Instead
After a DNS slip let a research model touch the live web, OpenAI moved frontier training into a cached copy of the internet. Here’s what changed.
