CVE-2026-53998: Phoenix: Format String Injection in applyChatTemplate
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 PHOENIXSECRET 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.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade to a fixed release to a version that resolves this vulnerability.
Fixed in 15.7.0
Event History
Frequently Asked Questions
Which deployments are exposed to unauthenticated exploitation?
Default Phoenix installations expose the affected applyChatTemplate GraphQL query without authentication. When authentication is enabled, any authenticated account can reach the query.
What can an attacker access through this issue?
A crafted F-string field can traverse Python object internals, including __class__, __init__, and __globals__, to read process environment variables and internal module state. Exposure of PHOENIX_SECRET may enable administrative session-token forgery and decryption of credentials protected by keys derived from that secret.
Which versions need remediation?
Phoenix versions from 13.0.0 until 15.7.0 are affected. The issue is fixed in version 15.7.0.