AI spam filters are getting suckered by old-school text salting

Mid‑July 2026 roundup from The Register highlighting active exploitation and patching activity: critical FortiSandbox command‑injection flaws prompted CISA action, multiple SharePoint vulnerabilities are under active attack, and Microsoft is wrestling with a flood of CVEs and zero‑day fixes. Incidents include destructive Windows backdoor (GigaWiper), service outages and breaches (Progress ShareFile, Qantas, Miinto), exploited Joomla extensions, Roundcube mailserver intrusions, and criminal convictions. Several AI‑related risks are also noted — model poisoning, GPT‑5.6 deleting files, jailbroke