Weekly Cyber Threat Roundup: Canvas Breach, AI Agent Hijacks, and Critical Patches

Top Attacks and Breaches

Instructure Canvas Platform Suffers Major Data Breach

Instructure, the U.S. education technology firm behind the widely used Canvas learning management system, has confirmed a significant data breach affecting its cloud-hosted environment. Threat actors accessed sensitive student and staff records, including private messages. The notorious group ShinyHunters escalated the incident by defacing hundreds of school login portals with ransom demands, amplifying the impact on educational institutions.

Zara Data Leak Tied to Third-Party Vendor

The flagship brand of Spanish fashion conglomerate Inditex, Zara, experienced a data breach linked to an external technology provider. Inditex acknowledged unauthorized access, and cybersecurity experts confirmed the exposure of 197,400 unique email addresses, order IDs, purchase histories, and customer support tickets. The incident highlights risks associated with third-party integrations in retail.

Hungarian Media Giant Mediaworks Hit by Extortion Attack

Mediaworks, which operates dozens of newspapers and online outlets in Hungary, fell victim to a data-theft extortion campaign. The breach came to light after the leak site World Leaks posted 8.5 TB of internal files online, including payroll records, contracts, financial documents, and internal communications. The company confirmed the intrusion as attackers used stolen data to pressure the organization.

Škoda Online Shop Compromised via Software Flaw

Czech automaker Škoda reported a security incident affecting its online shop. Attackers exploited a software vulnerability to gain unauthorized access. Exposed customer data may include names, contact details, order histories, and login credentials. The company assured that passwords and payment card data were not compromised in the incident.

AI Threats and Exploits

Critical WebSocket Hijack in Cline AI Coding Agent

Researchers discovered a critical WebSocket hijacking vulnerability in Cline’s local Kanban server, impacting the popular open-source AI coding agent. Rated CVSS 9.7, the flaw allowed any website a developer visited to exfiltrate workspace data and inject arbitrary commands into the AI agent. The issue was patched in version 0.1.66.

Anthropic’s Claude Chrome Extension Exposes Attack Surface

Security researchers identified a flaw in Anthropic’s Claude AI Chrome extension that enabled other browser extensions to hijack the AI assistant. Malicious prompts could trigger unauthorized actions and access sensitive browser-connected data. This vulnerability underscores how AI assistants expand the browser attack surface.

Fake Claude AI Installer Campaign Infects Users

An InstallFix campaign leveraged fake Claude AI installer pages promoted via Google Ads to infect Windows and macOS users. Victims were tricked into running commands that deployed multi-stage malware, stole browser data, disabled protections, and established persistence through scheduled tasks. The campaign demonstrates how threat actors exploit AI brand popularity.

Vulnerabilities and Patches

Progress MOVEit Automation Patches Critical Flaws

Progress has alerted customers to two vulnerabilities: CVE-2026-4670, a critical authentication bypass in MOVEit Automation allowing unauthorized access, and CVE-2026-5174, a privilege escalation flaw. Fixes are available in versions 2025.1.5, 2025.0.9, and 2024.1.8. Organizations using the managed file transfer software are urged to update immediately.

Ivanti Endpoint Manager Mobile Zero-Day Fixed

Ivanti has addressed CVE-2026-6973, a high-severity vulnerability in Endpoint Manager Mobile (EPMM) that was exploited as a zero-day. The flaw affects EPMM versions 12.8.0.0 and earlier, allowing attackers with administrator permissions to execute remote code. Hundreds of appliances remain at risk if patches are not applied promptly.