CVE-2026-53998

Phoenix: Format String Injection in applyChatTemplate

Published Nov 16, 2026 CVSS 3.1 · 10 Critical EPSS n/a KEV · Not listed
Description
Phoenix is an AI observability and evaluation platform. From 13.0.0 until 15.7.0, the applyChatTemplate GraphQL query passed attacker-controlled F-string templates to FStringTemplateFormatter._format, while DotDict.__getattr__ did not prevent traversal through Python private attributes. A crafted field could traverse object internals such as __class__, __init__, and __globals__ to read process environment variables and internal module state. The endpoint is unauthenticated on default installations, and any account can reach it when authentication is enabled. Exposure of PHOENIX_SECRET can allow administrative session-token forgery and decryption of credentials protected by keys derived from that secret. This issue is fixed in version 15.7.0.

Analysis

AI-generated

Technical impact

The vulnerability stems from a format string injection in the applyChatTemplate GraphQL query, where attacker‑controlled f‑string templates are passed to FStringTemplateFormatter._format. Because DotDict.__getattr__ does not block traversal to Python private attributes, an attacker can access internals such as __class__, __init__, and __globals__, allowing reading of environment variables, internal module state, and the PHOENIX_SECRET. The endpoint is exposed without authentication on default installations and is reachable by any authenticated account once authentication is enabled, enabling an adversary to read sensitive configuration and credentials and to forge administrative session tokens.

Affected Systems

Arize‑ai Phoenix versions from 13.0.0 through 15.7.0 are affected. The issue was fixed in release 15.7.0, so deployments running any earlier version that have not applied the upgrade are vulnerable.

Risk and Exploitability

With a CVSS score of 10 the vulnerability is rated critical; there is no EPSS score provided, indicating limited data on exploitation likelihood, yet the absence from KEV does not diminish the potential impact. The attack vector is remote over the network via the GraphQL endpoint, and requires only a crafted template payload. An unauthenticated attacker on a default install or any legitimate user on an authenticated system can exploit the flaw to read secrets, forge admin tokens, and potentially obtain full control of the Phoenix instance.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Upgrade Arize‑ai Phoenix to version 15.7.0 or later.
  • Restrict network access to the Phoenix GraphQL endpoint using firewall rules or network segmentation to block unauthenticated requests.
  • Disable or restrict the applyChatTemplate feature for unauthenticated users, or implement input validation to reject malicious f‑string templates.

Generated by OpenCVE AI on November 16, 2026 at 21:51 UTC.

Tracking

Sign in to view the affected projects.

Advisories

No advisories available
History
Type Values Removed Values Added
Description Phoenix is an AI observability and evaluation platform. From 13.0.0 until 15.7.0, the applyChatTemplate GraphQL query passed attacker-controlled F-string templates to FStringTemplateFormatter._format, while DotDict.__getattr__ did not prevent traversal through Python private attributes. A crafted field could traverse object internals such as __class__, __init__, and __globals__ to read process environment variables and internal module state. The endpoint is unauthenticated on default installations, and any account can reach it when authentication is enabled. Exposure of PHOENIX_SECRET can allow administrative session-token forgery and decryption of credentials protected by keys derived from that secret. This issue is fixed in version 15.7.0.
Title Phoenix: Format String Injection in applyChatTemplate
Weaknesses CWE-134
CWE-200
References
Metrics cvssV3_1

{'score': 10, 'vector': 'CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:L'}

Subscriptions

No data.

Weaknesses
  • CWE-134

    Use of Externally-Controlled Format String

  • CWE-200

    Exposure of Sensitive Information to an Unauthorized Actor

Sources

Status
PUBLISHED
Assigner
GitHub_M
Published
Updated
Nov 16, 2026, 7:50 PM
Reserved
Jun 11, 2026, 4:34 PM
Status
Received
Published
Nov 16, 2026, 8:19 PM
Modified
Nov 16, 2026, 8:19 PM