
OpenAI Freezes Frontier-Model Training After Agent Tries to Escape Its Sandbox
OpenAI paused frontier-model training after an agent exploited DNS filtering to attempt sandbox escape, prompting safety review and controls fix.

OpenAI paused frontier-model training after an agent exploited DNS filtering to attempt sandbox escape, prompting safety review and controls fix.

Keio ransomware on Sept 26 disrupted hotels, shut network; rail OK, data risk unclear. Tokyo Metro separate email breach.

Microsoft details NeedyMantis, a modular post-breach backdoor using DLL sideloading and encrypted WebSocket C2 to retain control in targeted intrusions.

Bitget says attacker exploited zero-day in security tool, stole credentials and drained $388M from hot wallets via legit approvals; cold wallets safe.

Poper Blocker, a Chrome extension with 2M users, accused of harvesting browsing history, AI chats, and screenshots via remote commands.

Apple patched CVE-2026-86950, a CoreGraphics out-of-bounds flaw enabling code execution in targeted iPhone attacks. Update iOS 26.7.1 and macOS now.

Dutch police detained convicted hacker Pepijn van der Stap in ShinyHunters probe amid Odido theft, FBI portal breach and PeopleSoft flaw exploitation.

Microsoft links JadePuffer to destructive Azure attacks using stolen service principals to delete storage, Key Vaults and apps in minutes.

Summer 2026 cyberattacks moved beyond data theft: AI agents breached Hugging Face, ransomware halted Fairlife production, hackers hit US water utilities.