External risk intelligence

Phoenix GraphQL Query F-string Vulnerability Allows Environment Variable Exposure

CVE advisorySeverity: CRITICAL (CVSS 10.0)

CVE-2026-53998

The vulnerability resides in a GraphQL endpoint of an AI observability platform. The endpoint is unauthenticated by default in standard installations and remains reachable by any user even when authentication is enabled, making it a public-facing service component that is accessible without prior authentication.

Information Disclosure

Halo Surface Signal: 5 out of 5 — more likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

An unauthenticated vulnerability in the Phoenix AI observability platform could allow unauthorized access to sensitive system information, potentially leading to session token forgery and the decryption of credentials. This issue affects versions prior to 15.7.0 and could have significant security implications if exploited.

  • Sensitive system data could be exposed.
  • Protects against unauthorized access to AI platform data.
  • Confirm platform relevance and exposure for security.

Attack Path

How an attacker could exploit the issue

An attacker could send a specially crafted request to the `applyChatTemplate` GraphQL query. This query, when processed by the platform, allows the attacker to influence an F-string template. The vulnerability lies in how this template is formatted, enabling an attacker to traverse object internals to access sensitive environment variables and module states. If successful, this could expose secrets like `PHOENIX_SECRET`, potentially leading to forged administrative sessions or decrypted credentials.

  • Unauthenticated network access required.
  • Triggers via crafted GraphQL query.
  • Secret exposure and session forgery.

Live Threat

Current exploitation, exposure, and threat context

A flaw in the applyChatTemplate GraphQL query could allow an attacker to read process environment variables, including sensitive secrets like PHOENIX_SECRET. When successful, this could lead to the forgery of administrative session tokens or the decryption of credentials, depending on the configuration and keys used.

  • Process environment variables and secrets.
  • Unauthenticated access to a GraphQL query.
  • Administrative session token forgery.

Operational Fix

Recommended remediation, mitigation, and detection steps

Application owners and platform teams are likely responsible for addressing this critical vulnerability in the Phoenix AI observability platform. The initial priority should be to identify all instances of the affected software, confirm their reachability and business criticality, and then assign an accountable owner for remediation planning.

  • Application or Platform teams own the issue.
  • Verify external reachability and business impact.
  • Plan phased remediation based on risk.

Supplementary metadata

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Frequently asked questions

What is the Phoenix platform?

Phoenix is an AI observability and evaluation platform designed to help teams monitor, troubleshoot, and optimize machine learning models and AI applications. It provides insights into how models perform by tracing interactions and evaluating model output quality, making it a central component for managing AI pipelines.

How does CVE-2026-53998 impact Phoenix?

This vulnerability involves an Improper Neutralization of Special Elements (CWE-134) and Information Exposure (CWE-200). Because the platform improperly processes user-supplied templates in a specific GraphQL query, it allows attackers to bypass security boundaries and read sensitive process memory, including environment variables like secret keys.

Do I need to trigger the vulnerability for it to work?

Yes, an attacker must send a specifically crafted request to the applyChatTemplate GraphQL endpoint. The vulnerability is not triggered by standard, legitimate use of the platform. However, simply navigating the interface or using the tool normally will not cause this issue to occur.

Why is this considered a high-priority risk?

Halo Surface Signal indicates this is highly relevant because the vulnerable GraphQL endpoint is unauthenticated by default. Even in installations where authentication is enabled, any user can reach the endpoint, making it accessible to unauthorized parties over the network.

When should I upgrade my Phoenix installation?

You should update to version 15.7.0 immediately. Upgrading is the primary way to remediate this flaw, as it contains the necessary security patches to fix how the platform handles template formatting and prevents unauthorized traversal of internal system objects.

References