Skip to main content
Mallory
HighPublic exploit

MLflow AI Gateway environment variable secret exfiltration

IdentifiersCVE-2026-4035CWE-201· Insertion of Sensitive Information…

CVE-2026-4035 affects mlflow/mlflow versions prior to 3.11.0. In vulnerable versions, the AI Gateway secret configuration permits the api_key field to contain $ENV_VAR references. At runtime, MLflow resolves those references against the MLflow server's own environment and then includes the resolved value in provider authentication headers sent to the configured upstream api_base. An attacker who can configure gateway secrets and direct requests to an attacker-controlled upstream endpoint can therefore cause MLflow to disclose sensitive server-side environment variables, including credentials such as AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY. The issue is exploitable by low-privileged authenticated users in deployments using basic-auth, and by unauthenticated users in default deployments without basic-auth.

Share:
For your environment

Are you exposed to this one?

Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.

ANALYST BRIEF

Impact, mitigation & remediation

What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.

Impact

What an attacker gets, and what they’ve been doing with it.

Successful exploitation results in exposure of sensitive server-side environment credentials to an attacker-controlled endpoint. Leaked credentials may include cloud artifact credentials and other secrets present in the MLflow server environment. This can enable follow-on compromise such as unauthorized access to external services, artifact poisoning, and cross-boundary code execution in downstream environments that consume attacker-modified artifacts or resources. The disclosed CVSS vector indicates high confidentiality impact with limited integrity and availability impact.

Mitigation

If you can’t patch tonight, do this now.

Until patched, restrict access to AI Gateway configuration and secret management to trusted administrators only. Enable and enforce strong authentication, including basic-auth where applicable, to prevent unauthenticated access present in default deployments without it. Prevent untrusted users from setting or modifying gateway secrets or upstream api_base endpoints. Where possible, avoid storing highly sensitive credentials in environment variables accessible to the MLflow server, and monitor outbound requests from MLflow for unexpected upstream destinations. If exposure is suspected, rotate affected credentials immediately.

Remediation

Patch, then assume compromise.

Upgrade mlflow/mlflow to version 3.11.0 or later, which fixes the environment variable resolution behavior in AI Gateway secrets. If managing source directly, apply the vendor fix associated with the referenced MLflow commit 4a3f2f720cb4f058c9e0c5b883e0acc9ab64a7f3. After upgrading, review AI Gateway configurations for untrusted upstream api_base values and rotate any environment-based credentials that may have been exposed.
PUBLIC EXPLOITS

Exploits

No public exploits tracked yet. Mallory keeps watching.

VALID 0 / 0 TOTALView more in app

No public exploit code observed for this vulnerability.

EXPOSURE SURFACE

Affected products & vendors

Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.

VendorProductType
LfprojectsMlflowapplication

Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.

What this page doesn’t show

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, which detections to deploy, and what to do tonight.
Exposure mapping

Query your assets running an affected version, and investigate the blast radius.

Threat actor evidence

Every observed campaign linking this CVE to a named adversary.

Associated malware

Malware families riding this exploit, with evidence and IOCs.

Detection signatures

YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.

Vendor-by-vendor mapping

Cross-references every affected SKU, including bundled OEM variants.

Social activity5

Community discussion across Reddit, Mastodon, and other social sources.