Affected Systems
MLflow versions < 3.15.0 (CVE-2026-64849, CVSS 9.3) and FUXA versions <= 1.2.9 (CVE-2026-25895, CVSS 9.5). MLflow is an open-source AI platform; FUXA is open-source SCADA/HMI software for industrial automation. Approximately 60 FUXA instances are publicly exposed.
Exploitation Status
Active exploitation confirmed. Attackers began scanning for vulnerable MLflow instances within hours of CVE assignment on August 17, 2026. Malicious scanning for FUXA instances detected starting August 18, 2026. MLflow attacks target cloud metadata services to exfiltrate credentials; FUXA attacks attempt file overwrites via path traversal, no RCE payloads observed yet.
Business Impact
Organizations running MLflow on cloud infrastructure face immediate risk of credential theft and secret exfiltration via SSRF attacks targeting internal metadata endpoints (e.g., AWS EC2 metadata, Azure IMDS). The flaw bypasses prior fixes due to improper handling of HTTP redirects in model-registry webhooks. FUXA users in OT/ICS environments risk arbitrary file writes and potential remote code execution, though active attacks currently focus on file overwrites. FUXA has a history of exploitation (CVE-2026-25939, CVE-2023-33831 exploited since November 2025).
Urgency
🔴 Immediate
Recommended Actions
- Upgrade MLflow to version 3.15.0 or later immediately for all internet-facing and cloud-hosted instances
- Review MLflow audit logs for suspicious HTTP requests to internal IP ranges (169.254.169.254, 100.64.0.0/10) and unusual model-registry webhook activity since August 17, 2026
- Rotate cloud credentials and secrets that may have been accessible from compromised MLflow servers, including IAM role credentials and service account tokens
- Upgrade FUXA to version 1.2.10 or later, or remove FUXA instances from public internet exposure and place behind VPN or firewall
- Monitor FUXA logs for unauthorized file write attempts targeting main.js or other application files, and review file integrity since August 18, 2026