‍

1. An AI Agent Just Ran a Ransomware Attack Without a Human

What happened

Researchers at Sysdig documented the first fully autonomous ransomware attack driven by an AI agent, tracked as JADEPUFFER. The agent exploited CVE-2025-3248, a critical unauthenticated remote code execution flaw in Langflow (a popular open-source framework for building LLM workflows), to gain initial access. From there it moved autonomously: harvesting credentials, pivoting to a production MySQL server running Alibaba Nacos, and encrypting 1,342 configuration records. It also repaired its own broken login in 31 seconds with no human involvement.

Why it matters

This is a meaningful shift. The attacker used the AI agent's ability to adapt mid-attack, diagnose failures, and self-correct to run what would normally require multiple operators. The skill floor for conducting a ransomware campaign has dropped significantly.

What to do

• Patch Langflow immediately to version 1.3.0 or later (CVE-2025-3248 is on CISA's Known Exploited Vulnerabilities list)
• Do not expose Langflow's code execution endpoints to the internet
• Review logs for self-correcting or self-modifying behaviour, which can now be an indicator of agentic attack activity rather than a routine user event

2. NetScaler Flaw Lets Attackers Walk Past MFA

What happened

CVE-2026-8451 is a pre-authentication memory overread vulnerability in NetScaler ADC and NetScaler Gateway appliances configured as a SAML identity provider. Rated CVSS 8.8, it allows an unauthenticated attacker to extract session tokens from memory via a malformed request, bypassing MFA entirely. Exploitation attempts were detected within 24 hours of Citrix disclosing the patch. This follows the same pattern as CitrixBleed (CVE-2023-4966) and CitrixBleed 2 (CVE-2025-5777), both of which were mass-exploited shortly after disclosure.

Why it matters

NetScaler sits in front of authentication for many enterprise environments. A session token extracted from memory gives an attacker authenticated access with no credentials needed. MFA provides no protection once the token is stolen.

What to do

• Patch NetScaler ADC and Gateway to the latest available version immediately
• Rotate active session tokens after patching
• Review SAML IDP logs for evidence of session hijacking or unusual authentication patterns

3. Ubiquiti UniFi Connect Scores a Perfect 10

What happened

CVE-2026-50746 is a CVSS 10.0 command injection vulnerability in the UniFi Connect Application (versions 3.4.16 and earlier). Any attacker with network access can send a single unauthenticated request to execute arbitrary commands on the host device. No login required. Approximately 100,000 UniFi OS endpoints are currently reachable from the public internet.

Why it matters

UniFi devices manage physical infrastructure including building access and networking. A full compromise gives an attacker a foothold that extends beyond the security perimeter into physical environments. This is the third critical advisory from Ubiquiti in under two months.

What to do

• Patch UniFi Connect Application to version 3.4.20 or later immediately
• UniFi Connect should not be on your core network if avoidable
• If remote access is required, place it behind a VPN

Stay safe and share this with your team.

Smarttech247 is a Gartner-recognised MDR provider. If any of these vulnerabilities affect your environment and you need support, get in touch.