Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

In early February 2026, users of Qinglong (青龙), a popular open source timed task management platform with over 19,000 GitHub stars, began reporting that their servers were maxing out CPU usage. The cause was a cryptominer binary called.fullgc, deployed through two authentication bypass vulnerabilities that allowed unauthenticated remote code execution. The attacks went largely unnoticed in the English-speaking security community.

On April 22, 2026, Google's Threat Intelligence Group and Mandiant disclosed a campaign by a threat actor they're tracking as UNC6692. The group breached enterprise networks by impersonating IT helpdesk staff over Microsoft Teams, ultimately exfiltrating Active Directory databases and achieving full domain compromise. What's notable about UNC6692 is what they didn't do. They didn't use a zero-day. They didn't exploit a software vulnerability.

A credential access event fired. An AI agent investigated it, correlated it against running processes, assessed the risk, and closed the ticket. No analyst touched it. The entire loop ran in minutes. This is what security operations look like when AI can actually operate in the environment rather than advise from outside it. Security operations have always required a special kind of person.

AI agents create data exfiltration risk by combining three capabilities that are dangerous together: access to private data, exposure to untrusted content, and the ability to communicate externally. When all three exist in one agent, an attacker can hide instructions inside an email, document, or webpage the agent processes and trick it into sending sensitive data out. No software vulnerability is required. The attacker doesn't need to break in. They just need to talk to your agent.

Last week, OpenAI released Privacy Filter, an open-weight model for detecting and redacting PII in text. It is a thoughtful release: Apache 2.0 licensed, able to run locally, designed for high-throughput workflows, and built to go beyond regex-based detection. This is good news for everyone building enterprise AI. Privacy at the model layer is getting real attention. What we liked most was how clearly OpenAI described the role of the model.

We're proud to announce that Frost & Sullivan has named CrowdStrike a Leader for the fourth consecutive time in the 2026 Radar for Cloud-Native Application Protection Platforms. This recognition validates our continued investment in combining posture management with real-time detection and response, and reinforces our leadership in stopping cloud attacks.

Understand where short-lived credentials reduce risk in agentic systems and where operational complexity requires stronger monitoring and governance controls.