Chrome Fixed 1,072 Bugs, Then Changed How Updates Reach Users
Chrome fixed more security bugs in two milestones than in the previous 23 while using AI across discovery, triage and fix preparation.
Read article →AI
AI systems, model security, agent abuse, and attacks that use generative tools.
Claude Breached Three Organizations During Anthropic Cyber Tests
A misconfigured evaluation harness let three Claude models reach production systems, exposing a control gap that prompts and model safeguards could not contain.
Read article →
Hermes Logs Record File Enumeration on a Thai Finance Ministry System
Five recovered task logs record service discovery, privilege checks and file enumeration after an operator enabled Hermes's unattended mode.
Read article →
OpenAI Cyber Test Breached Hugging Face Production
Hugging Face traced 17,600 actions from an Artifactory escape through two malicious-dataset vectors and into its clusters, network and source control.
Read article →
Crafted ChatGPT Link Created a Scheduled Workspace Agent in Tests
Zenity found URL parameters that preselected a template and auto-submitted instructions inside a logged-in user's Workspace Agents builder.
Read article →
CISA Confirms Exploitation of LiteLLM MCP Preview Command Injection
Any authenticated proxy-key holder could make two MCP preview endpoints run an arbitrary command on the LiteLLM host.
Read article →
Search Ads Used Real ChatGPT Pages to Deliver Malware
Push Security traced paid search ads to shared ChatGPT and Claude pages that handed visitors to fake desktop-app downloads.
Read article →
Google Sued a Phishing Service Behind 2.5 Million Scam Texts
A civil lawsuit targets a Telegram-based phishing-kit operation that Google links to 2.5 million texts and more than one million fraudulent URLs.
Read article →
GhostCommit Hid Secret-Theft Instructions Inside a PNG
Researchers put instructions to read .env and encode it into source inside a PNG that text-only pull-request reviewers ignored but vision-capable coding agents later followed.
Read article →
Fake Landing-Page Skill Reached 26,000 Agents After Passing Scanners
AIR says its brand-landingpage experiment reached 26,000 agents after scanners missed a remote instruction source that changed after approval.
Read article →
Amazon Q Ran MCP Commands Supplied by Project Repositories
CVE-2026-12957 allowed project configuration to start MCP processes with a developer's environment. AWS fixed the flaw in Language Servers for AWS 1.65.0.
Read article →
Gemini and ChatGPT Recommended Malicious GitHub Skills in Tests
Island found more than 800 fake skills and MCP servers using credible READMEs and ZIP files to turn capability searches into malware installs.
Read article →
Gemini CLI Rebuilt Botnet Command Infrastructure in Six Minutes
Trend Micro found a Russian-speaking actor using Gemini CLI to rebuild command-and-control infrastructure and operate eight infected computers.
Read article →Showing 14–26 of 30