CVE-2026-116690

LOWCVSS 7.5 · HIGH

Signal is active with 2 mentions in latest observed window

Immediate actions

  • Track advisory updates for patch or workaround availability

Recommended action window: Monitor and triage in normal cycle

NVD description

MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.10.0 until 3.17.0, the prompt branch in _create_model_version() accepts schemeless relative sources because path validation runs only when a source has a URL scheme, and mlflow.prompt.is_prompt is interpreted by tag presence rather than a true value. A remote caller can store a server-local path as the model version artifact location and use /api/2.0/mlflow/model-versions/get-artifact to read files accessible to the MLflow server process. Default unauthenticated servers require no privileges, while basic-auth deployments are exposed to authenticated users who can create registered models. This issue is fixed in version 3.17.0.

0.0/ 10 priority

Sources & remediation

Weakness type (CWE)
CWE-22

Priority

LOW

Exploitation

NONE

PoC

NONE

Patch

NONE

Momentum

NONE

Threat summary

  • 2 mentions across 1 observed day

What's happening

  • 2 total mentions across 1 day

Deep dive

Activity timeline2 mentions / 1d
01122Mentions · 2026-10-28: 210-28
Referenced assets1 URL
By indicator
Full discourse2 posts
  • Previdian@PrevidianCyber

    Over the past 24 hours, Previdian’s honeypot network has captured exploitation attempts targeting four recently disclosed vulnerabilities: 1. MLflow (CVE-2026-116690): Arbitrary file read through prompt model versions, which can expose sensitive files accessible to the MLflow server process. 2. Google Brotli (CVE-2026-111497): A decoder buffer overflow triggered by crafted compressed data, potentially allowing memory disclosure, denial of service or remote code execution. 3. Oracle E-Business Suite (CVE-2026-101767): Unauthenticated SQL injection in Application Object Library that can be chained to remote code execution. 4. Laravel (CVE-2026-114509): A TokenGuard authentication bypass that can enable account takeover under affected MySQL/MariaDB configurations. The attempts targeted our real application honeypots for MLflow, Brotli and Laravel, and our Oracle E-Business Suite decoy. All four have been added to Previdian’s Known Exploited Vulnerabilities (KEV) catalog, with supporting evidence and technical details. If these products are in your environment, review the affected versions and prioritize the relevant vendor fixes. Explore the evidence: http://previdian.com

    01121881
    314 followersView on X
  • Upwind Security MDR@UpwindMDR

    🚨 CRITICAL - RCE & Arbitrary File Read Vulnerabilities Patched in MLflow Two security vulnerabilities in MLflow allow unauthenticated attackers to compromise tracking servers: • CVE-2026-116692 (CVSS 9.8 Critical): Remote code execution via third-party scorer deserialization importing caller-placed artifact modules. • CVE-2026-116690 (CVSS 7.5 High): Path traversal in prompt model version source handling enabling arbitrary local file read. 👉 Affected: MLflow >= 3.10.0 / 3.12.0rc0 prior to 3.17.0 | Upgrade to v3.17.0 or later immediately

    0000023
    371 followersView on X

Explore more