Blog

OpenAI has deployed Control Room, a fail-closed system that can halt research-model output within 35 seconds after a safety or security alert.

Railway says an October breach exposed secrets in up to 100,000 workspaces, driving $740,000 in unauthorized AI usage and prompting refunds.

Mozilla patched five Firefox vulnerabilities, including a critical audio-video flaw rated CVSS 9.8, but reported no exploitation in the wild.

F-Secure’s new taxonomy maps 26 consumer harms across AI systems, highlighting risks from deceptive chatbots, deepfakes and hijacked autonomous agents.

A researcher bypassed Ray-Ban Meta glasses’ app-signature checks, exposing how paired-phone access could install unauthorized code on the wearable.

Adobe patched vulnerabilities across 15 products, including Acrobat, ColdFusion and Creative Cloud tools, with several flaws enabling arbitrary code execution.

SAP’s November security release fixes critical NetWeaver, Web Dispatcher and Gateway flaws, including unauthenticated vulnerabilities rated CVSS 10.0.

Google patches an actively exploited Chrome XML use-after-free flaw in version 157. Users on Windows, macOS and Linux should update promptly.

ESET says Russia-linked Turla used Cereal malware on government Exchange servers in Romania and Kazakhstan to intercept NATO and EU emails.