Top Cyber Threats This Week: Cisco 0-Day, Plugin4Shell RCE, Brevo ClickFix

Multiple cybersecurity incidents this week spotlight how attackers are exploiting trust, weak authentication, and plugin vulnerabilities to gain access, spread malware, or manipulate systems. Here’s a breakdown of the most urgent threats and lessons for orgs racing to keep up.

Main Threats Emerging

Cisco ISE Authentication Bypass (CVSS 10.0). Cisco disclosed a freshly discovered critical zero-day in its Identity Services Engine (ISE): CVE-2026-76460. An unauthenticated remote attacker can bypass authentication via an insecure API endpoint, gaining access to the management interface without valid credentials. The company confirmed active exploitation is underway. Patch immediately if you use Cisco ISE.

Plugin4Shell: AI Agent RCE across the board. Security researchers uncovered Plugin4Shell, a novel zero-click remote code execution flaw affecting AI coding agents including Claude Code, OpenAI Codex, GitHub Copilot, and Google Gemini CLI. The root issue: plugin SHA-pinning verification checks out the commit the market has pointed to without confirming that the content pulled really matches the expected code. An attacker who controls a plugin’s source repository can redirect it to serve malicious code while appearing to satisfy integrity checks. This bypass cannot be fully addressed via marketplace fixes—users must update their agents.

OpenAI’s linked vulnerabilities exploited via Claude. A chain of two security flaws—a misconfigured SSO gateway and an RCE bug in libheif—let attackers compromise OpenAI employees’ ChatGPT accounts and access private internal repositories. The libheif vulnerability (CVE-2026-32882) had been patched in version 1.22.0, and OpenAI moved quickly: the exploit was shut down within fourteen hours of disclosure.

KREMLIN malware targets browsers and credential theft. Threat analysts revealed an operation active since at least May 2025. Operators send fake banking prompts through malicious Chrome and Edge extensions to steal credentials, session tokens, and personal data. The actors use custom C++ installers, JavaScript loaders, and payloads impersonating Brazilian banks. The ecosystem is tracked under REF9334.

Incidents Impacting Consumers, Businesses

Brevo supply-chain “ClickFix” attack. More than 100,000 websites were compromised after an attacker abused a Cloudflare API key tied to Brevo. For roughly five and a half hours, malicious scripts were injected via Brevo’s own domain and customer JavaScript files. Visitors saw fake CAPTCHA prompts urging them to paste commands—an example of a ClickFix social engineering strategy that leads to malware. The attack also planted a WordPress plugin on customer sites and delivered overlays to site visitors.

Google Docs used to distribute Malware via custom panels. In another ClickFix scenario, hackers pretended to be crypto marketing execs and sent a Google Doc. The document included a sidebar (implemented via Apps Script) warning of a fake decryption failure. Following the prompts, victims downloaded malware (on macOS: AMOS infostealer; on Windows: a PowerShell loader chain). The sidebar also harvested IPs, geolocation, and crypto wallet info before triggering the payload.

Crypto wallets and Google Sheets abused for C2 and skimming. A still-active campaign is using Google’s Visualization API to run command-and-control via Sheets that host obfuscated JavaScript. Targets are tricked into copying scripts into their browser’s address bar or installing Tampermonkey extensions. The injected code skims data by swapping deposit addresses, changing clipboard contents, and overlaying fake UI elements. Attackers have confirmed at least 24 victims, funds stolen reaching ~$10,000 in just one campaign tracked.

Reappearance of Shai-Hulud worm in npm packages. Four npm packages tied to the worm surfaced again, despite a published hash match and registry scanning. The packages had been inactive for over 111 days before being resubmitted without being caught by registry security tools. The incident exposes gaps between registry scanning claims and real coverage.

Security Advancements & CVE Hot List

On the defense side, Google released AndroidX Security State v1.1.0 and its companion provider v1.0.0. Together, they offer developers a way to assess device security at a component level—moving beyond generic patch-level indicators to detect unpatched components and pending updates programmatically.

A new private preview from Google called Agent Anomaly Detection is emerging on the Gemini Enterprise Agent Platform. By analyzing reasoning traces, tool invocations, and execution flows, the system flags Deviations from expected behavior—an oversight layer aimed at catching misuse or violation by autonomous agents.

A long list of trending CVEs is also fueling urgency; notable among them are flaws in widely used tools and platforms (for example, CVE-2026-58138 for Orkes Conductor, a Nintendo Switch vulnerability, and multiple issues in Okta, Apache packages, Dell systems, and more). Organizations must triage and patch quickly.

Why this wave matters: attackers are exploiting small failures—broken defaults, unchecked permissions, trust in third-party code—to scale attacks widely. Relying on marketplace reviews, minimal checks, or assumed safe flows is no longer sufficient.

What to Watch

  • Monitor public vulnerability disclosures for rapidly developing exploits—especially for AI agents and trusted plugins.
  • Audit exposed APIs and authentication mechanisms—including behind dashboards, agent marketplaces, or management interfaces.
  • For supply-chain providers and CMS platforms, be vigilant about embedded code sources (widgets, plugins, extensions). Fake overlays or sidebars can be entry points.
  • Invest in detection that tracks behavior—not just signatures. Agent trace logging, reasoning chain analysis, and flow monitoring will matter more.

This week’s landscape underscores a simple truth: trust must be earned, not assumed. Even tools meant to protect, or ones long taken for granted, can turn into major threats when attackers find shortcuts in defaults, supply chains, or validation systems. Vigilance, quick patching, and control over what your tools can reach are proving essential in this escalating arms race.