Weekly Security Roundup — week ending 16 August 2026
from APJONE.UK Security Briefing
Security Stories — Week Ending 16 August 2026 Active Exploitation macOS Screen Sharing flaw under active exploitation — Dutch NCSC confirmed; attackers deploying Monero miners on internet-exposed systems with port 5900 open. Patch immediately. Max-severity SAP Commerce Cloud RCE exploited within 3 days of patch — Sets a benchmark: ~72 hours from disclosure to active exploitation for critical enterprise flaws. GeoServer zero-day under active exploitation (SecurityWeek) — SQL injection enabling RCE; no patch available. Apply network mitigations. Supply Chain ChainDrop npm worm evades standard defences (The Register) — New Shai-Hulud variant poisons hundreds of packages while leaving minimal traces in source repositories. Trivy, not LiteLLM, behind 2,500-org compromise (SecurityWeek) — The open-source vulnerability scanner was the actual attack vector; 95%+ of victims were compromised before malicious LiteLLM packages appeared. Breaches RingCentral: 1.6M accounts exposed by ShinyHunters (BleepingComputer) — Social engineering attack in July; PII now publicly dumped. French tax authority breach — 2M taxpayer records stolen (The Register) — Breach occurred June; advertised on criminal forum by alias ZeroBytes. Trezor: 14,000 customers' shipping data via ShipMonk breach (SecurityWeek) — Physical targeting risk for hardware wallet owners. 1,000+ UK charities hit by Beacon CRM breach (SecurityWeek) — Root cause: AWS key exposed in public JS build artifacts. Scottish Crown Office: 300 staff data exposed via supplier breach (The Register) APT & Nation-State HoneyMyte (Mustang Panda) adds kernel-level rootkit to CoolClient backdoor (Securelist) — Major capability upgrade for Chinese APT active across Asia. Apple sends new mercenary spyware threat notifications (BleepingComputer) — New wave of targeted spyware alerts; recipients should treat as confirmed compromise. Autonomous AI agents used to hack government and energy systems (The Register) — Open-source AI agents autonomously attacked critical infrastructure in early July. No longer theoretical. New Zealand NZSIS: China using space facilities for intelligence gathering (The Register) Key Themes Patch velocity: 72 hours or less from disclosure to active exploitation is now standard. Monthly patch cycles for internet-facing systems are no longer defensible. Supply chain depth: ChainDrop evades detection, Trivy as a vector, AWS keys in build artifacts — organisations are inheriting risk they cannot fully audit. Autonomous AI attacks: The shift from AI-assisted to fully autonomous offensive operations happened this week in documented incidents. Defensive frameworks have not kept pace. CISA KEV — No New Additions This Week Last entries: 11 August 2026 — CVE-2026-68820 (Windows kernel, Lazarus), CVE-2026-20349 (Cisco ASA), CVE-2026-72898 (Metabase RCE). View the full CISA KEV catalogue →
Embed this episode
What this episode covers
macOS exploitation, ChainDrop npm worm, RingCentral/French tax/Trezor breaches, HoneyMyte APT rootkit upgrade, and autonomous AI attacks on critical infrastructure.
No transcript yet.