GHSA-68G8-GH8F-7RHG
Vulnerability from github – Published: 2026-11-12 20:29 – Updated: 2026-11-12 20:29Impact
An unauthenticated remote caller could read part of the operator's configured LLM provider API key.
The query API relayed the text of a provider error straight to the caller. That text carried the upstream provider's own message, and providers characteristically quote a partially masked form of the credential they rejected when authentication fails — so a failing provider call published a fragment of the operator's key. A caller who can reach the endpoint can cause such a failure and read the result.
Who is affected: any deployment of pgEdge RAG Server at 1.0.0 or earlier, including all alpha and beta pre-releases, whose query endpoint was reachable by an untrusted party. The endpoint carries no authentication in the affected versions.
The disclosed material authenticates to a third-party provider account rather than to pgEdge software, which is why the severity carries Scope: Changed.
v1.0.0:internal/server/handlers.go:121 relayed err.Error() to the client when pipeline execution failed. parseError in
v1.0.0:internal/llm/openai/client.go produced that error by two routes, and both disclose:
var errResp ErrorResponse
if err := json.Unmarshal(body, &errResp); err != nil {
return fmt.Errorf("API error (status %d): %s", resp.StatusCode, string(body))
}
return fmt.Errorf("API error (status %d): %s", resp.StatusCode, errResp.Error.Message)
The first branch interpolates the entire unparsed response body. The second returns the provider's own message, which for OpenAI quotes the sk-proj- prefix and the final four characters of the submitted key.
Patches
Fixed in 2.0.0. Upgrade to 2.0.0 or later.
The fix adds internal/safeerr and stops provider error text reaching API clients.
Upgrading does not undo the disclosure, and this needs action beyond upgrading. Rotate any LLM provider credential that was configured on an affected version while the endpoint was reachable by an untrusted caller, particularly if that provider was rejecting the key at the time. A fragment already disclosed stays disclosed.
Workarounds
Place the server behind a proxy that rewrites 5xx response bodies, or terminate provider errors before they reach the API surface. Keeping provider credentials valid, so that no authentication failure is produced, also avoids the disclosure in practice.
None of these removes the underlying defect, and none is a substitute for upgrading.
{
"affected": [
{
"package": {
"ecosystem": "Go",
"name": "github.com/pgEdge/pgedge-rag-server"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "1.0.1-0.20260729022402-72e0785a6d7c"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-125326"
],
"database_specific": {
"cwe_ids": [
"CWE-209",
"CWE-522"
],
"github_reviewed": true,
"github_reviewed_at": "2026-11-12T20:29:32Z",
"nvd_published_at": "2026-11-10T18:17:34Z",
"severity": "MODERATE"
},
"details": "### Impact\n\nAn unauthenticated remote caller could read part of the operator\u0027s configured LLM provider API key.\n\nThe query API relayed the text of a provider error straight to the caller. That text carried the upstream provider\u0027s own message, and providers characteristically quote a partially masked form of the credential they rejected when authentication fails \u2014 so a failing provider call published a fragment of the operator\u0027s key. A caller who can reach the endpoint can cause such a failure and read the result.\n\n**Who is affected:** any deployment of pgEdge RAG Server at `1.0.0` or earlier, including all alpha and beta pre-releases, whose query endpoint was reachable by an untrusted party. The endpoint carries no authentication in the affected versions.\n\nThe disclosed material authenticates to a **third-party provider account** rather than to pgEdge software, which is why the severity carries Scope: Changed.\n\n`v1.0.0:internal/server/handlers.go:121` relayed `err.Error()` to the client when pipeline execution failed. `parseError` in\n`v1.0.0:internal/llm/openai/client.go` produced that error by two routes, and both disclose:\n\n```go\nvar errResp ErrorResponse\nif err := json.Unmarshal(body, \u0026errResp); err != nil {\n return fmt.Errorf(\"API error (status %d): %s\", resp.StatusCode, string(body))\n}\nreturn fmt.Errorf(\"API error (status %d): %s\", resp.StatusCode, errResp.Error.Message)\n```\n\nThe first branch interpolates the entire unparsed response body. The second returns the provider\u0027s own message, which for OpenAI quotes the `sk-proj-` prefix and the final four characters of the submitted key.\n\n### Patches\n\n**Fixed in `2.0.0`. Upgrade to `2.0.0` or later.**\n\nThe fix adds `internal/safeerr` and stops provider error text reaching API clients.\n\n**Upgrading does not undo the disclosure, and this needs action beyond upgrading. Rotate any LLM provider credential that was configured on an affected version while the endpoint was reachable by an untrusted caller**, particularly if that provider was rejecting the key at the time. A fragment already disclosed stays disclosed.\n\n### Workarounds\n\nPlace the server behind a proxy that rewrites 5xx response bodies, or terminate provider errors before they reach the API surface. Keeping provider credentials valid, so that no authentication failure is produced, also avoids the disclosure in practice.\n\nNone of these removes the underlying defect, and none is a substitute for upgrading.",
"id": "GHSA-68g8-gh8f-7rhg",
"modified": "2026-11-12T20:29:32Z",
"published": "2026-11-12T20:29:32Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/security/advisories/GHSA-68g8-gh8f-7rhg"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-125326"
},
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/pull/45"
},
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/commit/72e0785a6d7cc9e35ad67d570bc9fddeaa4b2f61"
},
{
"type": "PACKAGE",
"url": "https://github.com/pgEdge/pgedge-rag-server"
},
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/releases/tag/v2.0.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:L/I:N/A:N",
"type": "CVSS_V3"
}
],
"summary": "pgEdge RAG Server discloses part of the configured provider credential through API error messages"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.