blog

A critical code-injection vulnerability in GitLab, tracked as CVE-2026-19478 (CVSS 9.4), allows unauthenticated attackers to modify or delete publicly accessible projects and rewrite their data — with no credentials required. For any organisation hosting…

Researchers at Varonis disclosed CoSnitch, a flaw that let Microsoft Copilot “hack itself” and move data elsewhere without raising obvious red flags — reportedly the third critical exfiltration flaw found in Copilot this year.…

Adversa AI disclosed an attack on xAI’s Grok using “Cryptographic Context Injection” that can send a user’s name, location, subscription tier and conversation prompts to an attacker-controlled server. Guardrails and output handling are only…

As the Model Context Protocol (MCP) becomes the standard way to connect AI agents to tools and data, researchers are documenting tool-poisoning attacks — malicious instructions hidden in tool descriptions the model reads but…

Attackers are exploiting CVE-2026-59310, a severe directory-traversal vulnerability in VMware vCenter, to deploy Babuk-derived ransomware. Internet-exposed management interfaces remain a favourite ransomware entry point. The gap between “patch available” and “patched everywhere” is precisely…

An extortion group called Icarus reportedly broke into Klue’s systems using a credential issued back in 2022, exposing keys to customers’ cloud services and enabling data theft for extortion. In the cloud, identity is…

Microsoft researchers found a path in Semantic Kernel where a single prompt could turn prompt injection into host-level remote code execution — code running on the device hosting the AI agent. Autonomous agents that…

Microsoft’s August 2026 Patch Tuesday addressed hundreds of CVEs, including CVE-2026-68820 — an actively exploited elevation-of-privilege zero-day in the Windows Ancillary Function Driver for WinSock. Patching matters. The more useful question is what an…