Tag
ai-security
41 posts with this tag.

·7 min read
Vercel agent-desktop: What an AI Agent Can Read
I ran agent-desktop 1.2.0 on Ubuntu with a fake password open in a second app. It read the password with no prompt. What it blocks, stores and sends.

·8 min read
Ollama Security: Check If Your AI Server Is Exposed
Ollama's local API has no login, and an attacker in Japan listed AI servers as targets. How to check yours, the Docker trap, and the fixes.
·7 min read
ARTEX AI Hacking Arrests in Korea: Extortion Playbook
Seoul police arrested 7 over AI-assisted hacks of 90 firms. 75 victims never reported. Six steps for a small team that gets the extortion email.
·7 min read
AGENTS.md Security: The Downloads Folder Attack on Codex
A planted AGENTS.md in ~/Downloads got Codex to install malware in tests. Six steps for teams on Codex or Claude Code, plus what to log.
·7 min read
Unauthorized AI Agents in Bank Accounts: Who Pays?
US banks are blocking AI agents that log in with customers' passwords. Chase's new terms and Regulation E put the loss on whoever shared the login. What to do.

·6 min read
ConnectWise Cloud Backup Breach: What to Check in M365
Microsoft tied the ConnectWise Cloud Backup (SkyKick) breach to Storm-0940 and an AI attack tool. The app IDs to find in Entra and what to check next.
·7 min read
LiteLLM 1.107.0 malware and Railway leak: keys to rotate
LiteLLM 1.107.0 shipped a credential stealer and Railway says env vars may have leaked. The AI keys to rotate, plus the PHP, GitLab and Chrome patches.
·5 min read
Meta Muse Privacy Leak: What to Do if Your Team Uses It
Did the Meta Muse privacy leak expose your team's chat titles? Shared links sat in public DNS records. See what Meta paused and what to do now.
·9 min read
Deepfake CEO Fraud: The Fideuram Case and How to Stop It
A fake WhatsApp from the CEO and an AI-cloned lawyer voice moved Fideuram money abroad in hours. What happened and the payment check that stops it.

·7 min read
BC Sues OpenAI Over Shooting Chat Logs: What to Check
British Columbia sued OpenAI over Tumbler Ridge. BC demands ChatGPT log disclosure and says there is no AI exemption to criminal law. 3 enterprise checks.
·5 min read
Meta Muse Zero-Day: AI Agent Device Access Vetting
Meta Muse 0-day lets local malware hijack the agent account via ClickFix. Amazon blocked Muse. 3 checks before approving AI assistants with device access.
Showing 12 of 41 posts.