This briefing covers 20 cybersecurity and geopolitics
stories published around Monday, November 16, 2026,
and 4 disclosed vulnerabilities
(CVE-2026-129952, CVE-2026-129955, CVE-2026-129956, CVE-2026-129957).
Each entry links to the original reporting.
Google GTIG and Mandiant report threat actors have moved beyond basic AI prompting to deploying autonomous multi-agent frameworks, automated vulnerability exploitation pipelines, and LLMJacking monetization schemes.
Kaspersky ICS CERT attributed a months-long compromise of a Russian software developer's build and code-signing infrastructure to BO Team (aka Black Owl, Hoody Hyena), entry gained via Log4Shell in VMware vCenter.
Chinese threat actors breached Barracuda while broadly adopting the Artex agentic AI penetration-testing framework to accelerate offensive operations. Operationalizing AI-driven attack automation marks a capability step-change that compresses defenders' detection-to-patch windows.
Bahrain's military intercepted multiple Iranian aerial threats as Iran's IRGC claimed it struck a radar system in the region. The exchange signals active escalation between Iran and Gulf states amid broader regional conflict dynamics.
A November 10 attack forced Lithuania's customs agency to shut down its main user management portal and triggered a possible data breach. Targeting a EU border agency handling trade and traveler data carries supply-chain and intelligence implications for the broader Baltic region.
Manila and Canberra signed a new Defense Cooperation Agreement expanding on the 2007 Status of Visiting Forces Agreement. The accord deepens the anti-China alignment arc across Southeast Asia and signals Canberra's growing willingness to anchor security commitments closer to the South China Sea.
Russia emerged from recent diplomacy with implicit U.S. willingness to discuss the "European security order"—Kremlin code for recognized spheres of influence over Eastern Europe. The concession, if formalized, would validate Moscow's core strategic demand and undercut NATO's open-door principle without a shot fired.
Washington and allied governments deepened a supply-chain security agreement explicitly targeting 'hostile actors' — a thinly veiled effort to exclude Chinese components from AI infrastructure. The deal doubles as economic coercion, pressuring partner nations to align with US tech-decoupling strategy against Beijing.
The UK National Resilience Committee found the forthcoming Cyber Security and Resilience Bill inadequate against state-level threats, publishing the rebuke eight days before Parliament votes.
The EU sanctioned a pro-Russian hacktivist group and 24 employees of the Moscow firm running the Doppelgänger disinformation campaign. The move names and freezes assets of operational personnel, raising the cost of Kremlin-linked influence operations targeting European audiences.
Financially motivated cluster REF3364 deployed SANSAR, a Rust-based Linux cryptomining orchestrator using TOR for C2, that tampers with NVIDIA GPU telemetry to evade detection. The technique defeats GPU-load-based detection controls, undermining a primary defense used by cloud and enterprise security teams.
An attacker compromised Fireworks AI, stole an access key, exfiltrated credentials, and used them to push malicious copies of the Claude Code extension into the Cursor IDE ecosystem. The incident illustrates how AI toolchain providers have become high-value pivot points for downstream developer supply-chain attacks.
Kaspersky identified a new CoffeeLoader variant using a resurrected Retadup worm infrastructure, dropping a patched 7z.dll via rundll32.exe to execute payloads. Reanimating a dismantled botnet's distribution chain signals adversaries are recycling law-enforcement-disrupted infrastructure rather than building fresh.
Phishing emails impersonating Microsoft 365 service notices leveraged a genuine outage to threaten message deletion and harvest credentials. Opportunistic abuse of real infrastructure events accelerates victim trust collapse and shortens the social-engineering kill chain.
Dutch firm Midnight Blue found six unpatched flaws in Danfoss AK-SM800A refrigeration controllers allowing unauthenticated remote takeover; firmware 5.4.0 does not fix them.
A Google AI agent independently analyzed a security patch and generated a functional exploit with zero human involvement. Fully autonomous exploit development eliminates the skilled-researcher bottleneck, threatening to democratize zero-day weaponization at machine speed.
CVE-2026-129957 (CVSS 8.x) lets authenticated PrestaShop employees write arbitrary PHP files to web-accessible paths via crafted theme or module archives in versions before 8.2.9 and 9.2.1. Any compromised or malicious back-office account becomes a full remote code execution vector on affected storefronts.
CVE-2026-129956 (CVSS 8.5 HIGH) in Sandboxie-Plus before 1.18.6 lets standard users escalate to SYSTEM by injecting code into a signed SandMan.exe process and triggering MSGID_PROCESS_RUN_UPDATER. A sandboxing tool delivering privilege escalation to SYSTEM is a direct contradiction of its security guarantee.
CVE-2026-129955 (CVSS 9.8 CRITICAL) allows unauthenticated attackers to gain full NodePress admin control using the hardcoded default password 'root' when NP_ADMIN_DEFAULT_PASSWORD is unset. Any internet-exposed NodePress instance running through 8.5.1 is trivially ownable with no prior access.
CVE-2026-129952 allows unauthenticated attackers to forge HS256 admin JWTs using NodePress's hardcoded 'nodepress-secret' fallback key, granting access to all administrator APIs through version 8.5.1.