AI agents behind ConnectWise breach, PHP2Shell exploited days before patches, Japanese cloud hack

PHP Exploited Before Patch, Unpatched Linux SMB Bugs, GitLab Max-Severity Fix, and AI Agents in ConnectWise & Japan Breaches Cybersecurity Today (Oct. 23, 2026) covers active exploitation of two critical PHP RCE flaws beginning at least four days before patches shipped, with multiple firms seeing failed pre-disclosure attempts and the UK NCSC urging immediate patching or taking servers offline. It also details two publicly disclosed, unpatched Linux kernel ksmbd SMB RCE bugs (one unauthenticated on port 445), exposed after ZDI credited Trend Micro's AI scanner and copied public mailing lists without patches. GitLab issued fixes for a maximum-severity unauthenticated RCE (CVE-2026-11-0264) in its regex compiler, following recent high-severity GitLab issues. Microsoft attributed the ConnectWise cloud backup compromise to China-aligned Storm-0940 using the Artex agentic AI framework, and Artex-related activity was also found in attacks on Japanese companies amid a broader breach wave, including a malware incident affecting up to 26,739 Haname Desktopcloud customer servers. 00:00 Headlines And Intro 00:33 PHP To Shell Exploitation 02:07 Patch Now Impact Scope 02:47 Linux KSMBD Bugs Exposed 04:10 AI Disclosure Process Drama 05:25 GitLab Max Severity Patch 07:30 ConnectWise Breach AI Agents 09:24 Artex Spreads In Japan 11:18 GMO Haname VPS Malware 12:24 Closing And Sign Off