Cybersecurity
Feb 3, 2026
Mar 3, 2026
•
6 min read
.png)
Cybersecurity threats continued to escalate this week across industries and geographies. A maximum-severity Cisco zero-day drew a rare Five Eyes joint advisory, a Chinese espionage campaign was unraveled across 42 countries, and a novel attack on AI developer tooling raised urgent new questions about emerging attack surfaces.
Below is a concise roundup of the most important cybersecurity developments from the past seven days — what happened, and why it matters.
Explore this content with AI:
ChatGPT | Perplexity | Claude | Google AI Mode | Grok
Cisco SD-WAN zero-day has been actively exploited since 2023
Security agencies from the U.S., UK, Canada, Australia, and New Zealand jointly sounded the alarm this week over a newly disclosed maximum-severity vulnerability in Cisco's Catalyst SD-WAN Controller and SD-WAN Manager products. With a perfect CVSS score of 10.0, the flaw lets an unauthenticated remote attacker bypass authentication entirely and gain administrative access to an organization's network.
What makes this especially alarming: investigations suggest the vulnerability has been actively exploited in the wild since at least 2023 — meaning attackers may have had quiet, persistent access to affected networks for years before detection. Organizations running these products should treat patching as an emergency priority.
Why it matters: SD-WAN infrastructure underpins wide-area networking for enterprises globally. A bypass at this layer can hand attackers complete visibility into — and control over — an organization's traffic routing.
APT28 weaponizes Microsoft MSHTML zero-day
Russia-linked threat actor APT28 — also known as Fancy Bear — was linked this week to active exploitation of a high-severity zero-day in Microsoft's MSHTML Framework. The flaw, rated 8.8 on the CVSS scale, allows an unauthenticated attacker to bypass security features entirely. Microsoft patched it in February's Patch Tuesday update but simultaneously confirmed it had already been weaponized in the wild — sending IT teams scrambling to verify their patch status.
MSHTML is the underlying rendering engine used by numerous Windows components and legacy web-based applications, meaning the attack surface is exceptionally broad.
Why it matters: APT28 consistently targets government agencies, defense contractors, and critical infrastructure. A weaponized MSHTML flaw in their hands is a stealthy, high-value intrusion vector that bypasses many conventional endpoint defenses.
Google disrupts Chinese GridTide campaign spanning 53 organizations in 42 countries
Google's Threat Intelligence Group announced it disrupted a sophisticated Chinese cyberespionage campaign that quietly compromised 53 organizations across 42 countries. The operation deployed a novel backdoor malware dubbed GridTide, which executed shell commands and exfiltrated files while disguising its activity as routine API traffic — an evasion technique that rendered the intrusions nearly invisible to standard network monitoring.
Targeted sectors included telecommunications providers, government ministries, SaaS platforms, and large enterprises across the Americas, Asia, and Africa. Attackers used cloud-hosted command-and-control infrastructure, complicating both attribution and takedown efforts.
Why it matters: The GridTide campaign is a prime example of "living off trusted services" — using legitimate cloud infrastructure so malicious traffic blends in with normal business operations, defeating perimeter defenses entirely.
Madison Square Garden confirms breach from Cl0p's Oracle attack
Madison Square Garden Entertainment this week formally confirmed it was a victim of the Cl0p ransomware group's mass exploitation of Oracle's E-Business Suite — a campaign that ultimately hit more than 100 organizations. MSG was first named as a victim in November 2025, when over 210GB of archived files were leaked after the company reportedly declined to pay a ransom.
MSG has now begun notifying individuals whose personal data was exposed. The Oracle EBS instance involved was managed by a third-party vendor — making this yet another case where a breach originated outside a company's own walls.
Why it matters: Third-party and vendor risk remains one of the most underappreciated attack surfaces. Organizations frequently discover they have been breached not through their own systems, but through a contractor trusted to handle sensitive data on their behalf.
ClawJacked: Flaw in AI agent platform OpenClaw allowed silent browser takeover
Researchers at Oasis Security disclosed a high-severity flaw in OpenClaw, a popular AI agent platform — dubbed ClawJacked. The vulnerability lives in OpenClaw's core gateway and allows a malicious website to connect to a locally running AI agent instance and silently take control of it, without the developer's awareness or consent.
The attack requires only that a developer be running OpenClaw locally and visit a compromised site. The site can then brute-force access to the local WebSocket server and inject commands for the AI agent to execute. OpenClaw has since issued a patch, but the disclosure raises urgent questions as AI agents become standard in development workflows.
Why it matters: AI agents increasingly have access to files, code repositories, cloud credentials, and APIs. Vulnerabilities in their runtimes represent a new class of high-value target — with a much larger blast radius than traditional endpoint compromise.
New Zealand medical platform MediMap hacked — patient records altered
New Zealand medical records platform MediMap was compromised by an unknown threat actor who didn't just access patient data — they actively modified records, with some entries defaced and others reportedly marked as deceased. MediMap is used by pharmacies and healthcare providers to communicate critical medication dosage information, making record tampering a direct patient safety issue.
The incident remains under investigation. New Zealand health authorities are working with MediMap to identify which records were affected and whether any clinical decisions may have been influenced by falsified data.
Why it matters: Most data breaches involve theft — this one involved manipulation. Integrity attacks on healthcare platforms represent a more dangerous threat vector than exfiltration, with real potential for patient harm from falsified records.
Read more at 1News New Zealand
Final thoughts
This week's headlines reinforce a recurring theme: attackers are patient, diverse in their methods, and increasingly skilled at hiding in plain sight. A Cisco zero-day exploited undetected for three years, AI agents hijacked through the browser, and Chinese espionage malware disguised as API traffic all point to the same strategic reality — perimeter defenses alone are no longer enough.
For security teams, the takeaways are familiar but urgent: rapid patch management, zero-trust architecture, rigorous third-party risk assessments, and continuous monitoring — especially across AI tooling now deeply embedded in development environments. Check back next week for another roundup of the cybersecurity stories shaping the threat landscape.