Table of Contents
Two critical vulnerabilities are under active exploitation right now — one in MLflow, the open-source AI platform used by thousands of organizations for machine learning operations, and one in FUXA, a web-based SCADA system used in industrial automation. The MLflow vulnerability, tracked as CVE-2026-64849, has a CVSS score of 9.3 and allows unauthenticated attackers to access internal cloud metadata endpoints and extract sensitive data. The FUXA vulnerability, tracked as CVE-2026-25895, has a CVSS score of 9.5 and allows unauthenticated remote attackers to write arbitrary files to the server and achieve remote code execution. Both vulnerabilities were documented by watchTowr and VulnCheck in August 2026, with active scanning and exploitation efforts confirmed in the wild. For any organization running MLflow for AI model tracking or FUXA for industrial control systems, these are patch-now vulnerabilities — not patch-eventually. For Filipino IT professionals, cybersecurity teams, and developers using open-source AI and OT platforms, the simultaneous exploitation of AI infrastructure and industrial control vulnerabilities signals a convergence of threats that traditional security tools are not designed to handle.
Key Takeaway
- 🔴 CVE-2026-64849 (MLflow): CVSS 9.3 — Unauthenticated Server-Side Request Forgery (SSRF) vulnerability in MLflow’s Tracking Server. Allows attackers to access internal cloud metadata endpoints and extract credentials. Affects versions before 3.15.0. Under active exploitation.
- 🔴 CVE-2026-25895 (FUXA): CVSS 9.5 — Missing authentication and path traversal vulnerability in FUXA SCADA software. Allows unauthenticated remote attackers to write arbitrary files and achieve remote code execution. Affects industrial control systems.
- ⚡ Both Under Active Attack: watchTowr and VulnCheck confirmed malicious scanning and exploitation efforts for both vulnerabilities in August 2026. This is not theoretical — attackers are actively probing for exposed instances.
- 🇵🇭 Filipino Impact: MLflow is used by Filipino AI developers and BPO companies for model tracking. FUXA is used in industrial automation and OT environments. The Philippines’ 100% supply chain cyber incident rate means any exposed instance is a potential entry point.
- 🔧 Action Required: Patch MLflow to version 3.15.0 or later. Update FUXA to the latest version. If patching is not immediately possible, restrict network access to both systems and block external internet exposure.
CVE-2026-64849: The MLflow SSRF Vulnerability
MLflow is an open-source platform for managing the machine learning lifecycle — experiment tracking, model registry, and deployment. It is used by thousands of organizations worldwide, including many in the Philippines’ growing AI and BPO sectors. As The Hacker News reported, CVE-2026-64849 is an unauthenticated Server-Side Request Forgery (SSRF) vulnerability in MLflow’s Tracking Server.
SSRF vulnerabilities allow an attacker to make the vulnerable server send requests to internal resources that are not accessible from the outside. In the case of MLflow, the vulnerability allows an attacker who can reach the Tracking Server to issue HTTP requests to arbitrary internal cloud metadata endpoints — the endpoints that cloud providers (AWS, Azure, GCP) use to deliver temporary credentials to virtual machines. By accessing these endpoints, an attacker can extract cloud credentials, API keys, and other sensitive data that the MLflow server has access to. As OWASP’s SSRF documentation explains, this class of vulnerability is particularly dangerous in cloud environments where metadata endpoints provide direct access to IAM credentials.
The vulnerability affects MLflow versions before 3.15.0. The fix is to upgrade to version 3.15.0 or later, which was released to address this issue. As watchTowr documented, the vulnerability is particularly dangerous in cloud environments where MLflow is deployed on virtual machines with IAM roles — the attacker can steal the IAM credentials and gain access to the entire cloud environment. For more on cloud security for Filipino organizations, see our AI agent security guide.
CVE-2026-25895: The FUXA SCADA Vulnerability
FUXA is an open-source, web-based SCADA (Supervisory Control and Data Acquisition) and HMI (Human-Machine Interface) software designed for operational technology (OT) and industrial automation. It is used to monitor and control industrial processes — manufacturing lines, power plants, water treatment facilities, and building automation systems. CVE-2026-25895 combines two critical vulnerabilities: missing authentication for a critical function and a path traversal vulnerability.
The missing authentication component means that an unauthenticated remote attacker can access a function that should require login. The path traversal component means the attacker can write arbitrary files to the server’s file system. Combined, these vulnerabilities allow an attacker to write a malicious file — such as a web shell or a reverse shell script — to the server and execute it, achieving remote code execution. This gives the attacker full control of the FUXA server and, potentially, the industrial processes it controls.
The implications for industrial environments are severe. If an attacker gains control of a FUXA SCADA system, they can manipulate industrial processes, cause physical damage, or use the compromised system as a pivot point to attack the broader OT network. For more on industrial cybersecurity, see our coverage of AI agent security risks and our supply chain attack guide.
Why Both Vulnerabilities Matter for Filipino Organizations
The simultaneous exploitation of MLflow and FUXA vulnerabilities highlights a convergence of threats that Filipino organizations need to understand:
1. AI infrastructure is now a target. MLflow is not a traditional IT system — it is an AI platform. The fact that attackers are actively exploiting it means that AI infrastructure is now in the crosshairs. Filipino BPO companies and AI startups that use MLflow for model tracking are potential targets. The vulnerability allows attackers to steal cloud credentials, which can lead to data breaches and ransomware. For more on how AI is changing the threat landscape, see our AI cyberattacks analysis.
2. Industrial control systems are exposed. FUXA is used in OT environments that many Filipino organizations do not consider part of their IT security scope. The convergence of IT and OT means that a vulnerability in a SCADA system can be the entry point for a broader attack. Filipino manufacturing companies, utilities, and infrastructure operators should audit their FUXA deployments immediately. For practical guidance, see our cybersecurity Philippines guide.
3. The 100% supply chain problem. BlueVoyant’s 2025 report found that 100% of surveyed Philippine organizations experienced cybersecurity incidents linked to supply chain vulnerabilities. MLflow and FUXA are both open-source projects — part of the software supply chain. When these projects have vulnerabilities, every organization using them is exposed. Filipino organizations need to track vulnerabilities in their open-source dependencies, not just their commercial software. See our supply chain attack defense guide for how to manage this risk.
How to Protect Your Organization
For Filipino IT and security teams, the response to these two vulnerabilities should be immediate:
- Patch MLflow immediately. Upgrade to version 3.15.0 or later. If you cannot patch immediately, restrict network access to the MLflow Tracking Server — it should never be exposed to the internet. Use firewall rules to limit access to authorized users only.
- Patch or isolate FUXA. Update to the latest version. If patching is not possible, isolate the FUXA server on a segmented network with no internet access. SCADA and OT systems should never be reachable from the public internet.
- Audit cloud credentials. If you run MLflow in AWS, Azure, or GCP, check your IAM logs for unusual activity. If the Tracking Server was exposed, assume credentials were stolen and rotate them immediately.
- Scan for exposed instances. Use tools like Shodan or Censys to check whether your MLflow or FUXA instances are publicly accessible. If they are, close the exposure immediately.
The Convergence of AI and OT Security Threats
The simultaneous exploitation of mlflow and FUXA vulnerabilities represents a new pattern in cybersecurity: the convergence of AI infrastructure threats and operational technology threats. Historically, AI platforms like mlflow were managed by data science teams, while SCADA systems like FUXA were managed by engineering teams. Security teams rarely overlapped. But as AI models are increasingly deployed in industrial environments — for predictive maintenance, process optimization, and autonomous control — the boundaries between AI infrastructure and OT infrastructure are blurring. A vulnerability in mlflow can now be the entry point to an OT network, and a vulnerability in FUXA can be the entry point to an AI infrastructure. Filipino organizations need security teams that understand both domains. For more on how to build this capability, see our guide to AI skills for Filipino professionals and our AI cyberattacks analysis.
The mlflow vulnerability also highlights a broader risk in the AI supply chain. Open-source AI tools — mlflow, OpenClaw, and hundreds of others — are now critical infrastructure for thousands of organizations. But open-source projects often lack the resources for continuous security auditing. When a vulnerability is found, every organization using the tool is exposed simultaneously. This is the same supply chain risk pattern we documented in our analysis of the OpenClaw security crisis, where 42,665 exposed instances were found vulnerable. The lesson for Filipino organizations is clear: track your open-source dependencies, monitor for CVEs, and have a patching plan ready before a vulnerability is exploited. The mlflow SSRF vulnerability is a reminder that the AI tools you rely on for model development are also attack surfaces that adversaries actively target.
Frequently Asked Questions
What is CVE-2026-64849?
CVE-2026-64849 is a critical (CVSS 9.3) SSRF vulnerability in MLflow, an open-source AI model tracking platform. It allows unauthenticated attackers to access internal cloud metadata endpoints and extract credentials. It affects MLflow versions before 3.15.0 and is under active exploitation as of August 2026.
What is CVE-2026-25895?
CVE-2026-25895 is a critical (CVSS 9.5) vulnerability in FUXA, an open-source SCADA software used in industrial automation. It combines missing authentication with path traversal, allowing unauthenticated remote attackers to write arbitrary files and achieve remote code execution on industrial control systems.
Are these vulnerabilities being exploited?
Yes. Both watchTowr and VulnCheck confirmed active scanning and exploitation efforts for both vulnerabilities in August 2026. Attackers are actively probing for exposed MLflow and FUXA instances on the internet.
How do I fix the MLflow vulnerability?
Upgrade MLflow to version 3.15.0 or later. If you cannot patch immediately, restrict network access to the MLflow Tracking Server so it is not exposed to the internet. Check cloud IAM logs for unusual activity if the server was exposed.
How do I fix the FUXA vulnerability?
Update FUXA to the latest version. If patching is not immediately possible, isolate the FUXA server on a segmented network with no internet access. SCADA and OT systems should never be publicly accessible. See our cybersecurity Philippines guide for more.
How do these vulnerabilities affect Filipino organizations?
Filipino BPO companies and AI startups using MLflow are at risk of credential theft. Filipino manufacturing and utilities using FUXA are at risk of industrial system compromise. With 100% of Philippine organizations experiencing supply chain cyber incidents, open-source vulnerabilities like these are a critical threat vector. See our supply chain attack defense guide.
