CVE-2026-117574

validating

no check availablenot proven exploitable

cve.orgpublic pockev

Red Hat OpenShift AI (RHOAI)

CWE-639 Authorization Bypass Through User-Controlled Key

ETK-2026-2901, permanent identifier

cvss9.1cna vendorCVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:L/A:L

–running

startingcurrent stage

timeline

0 = trigger signal

trigger signal (cve.org) +0h

+2h

+4h

+6h

+8h

+10h

+12h

description

ai-gateway-payload-processing can use another tenant's API credentials when an authenticated caller selects that tenant's model. An attacker with gateway access may make unauthorized model requests and deplete another tenant's LLM quota; requests must be able to specify a cross-tenant model identifier.

Still under analysis: more detail becomes available as the pipeline runs.

mitigation

Restrict gateway access to trusted tenant-specific identities and block requests that select models outside the caller's tenant. Enforce tenant-to-model authorization at an upstream gateway or proxy until an updated package is provided.

check

building now

exploit

not yet

automation

–

auth required

yes, low privileges

EPSS

–

agent run time

–

cve.org record, t = 0

2026-10-28 20:02Z

Δ to public poc

–

Δ to kev

–

outcome

Run in progress: no artifact yet.

source signals

Public signals that triggered or updated the pipeline. These are inputs for the deltas above.

2026-10-28 20:02Z
cve_org: trigger signal, started the run
https://www.cve.org/CVERecord?id=CVE-2026-117574

this entry in the feed

GET /api/v1/feed/CVE-2026-117574

{
  "etk_id": "ETK-2026-2901",
  "cve_id": "CVE-2026-117574",
  "display_id": "CVE-2026-117574",
  "ghsa_id": null,
  "euvd_id": null,
  "status": "in_pipeline",
  "anchor_state": "anchored",
  "title": "Cross-tenant authorization bypass in ai-gateway-payload-processing",
  "summary": "ai-gateway-payload-processing can use another tenant's API credentials when an authenticated caller selects that tenant's model. An attacker with gateway access may make unauthorized model requests and deplete another tenant's LLM quota; requests must be able to specify a cross-tenant model identifier.",
  "mitigation": "Restrict gateway access to trusted tenant-specific identities and block requests that select models outside the caller's tenant. Enforce tenant-to-model authorization at an upstream gateway or proxy until an updated package is provided.",
  "vendor": "Red Hat",
  "product": "Red Hat OpenShift AI (RHOAI)",
  "affected": [
    {
      "affected_notes": [],
      "affected_ranges": [],
      "ecosystem": null,
      "fixed_versions": [],
      "package_name": "rhoai/odh-ai-gateway-payload-processing-e2e-rhel9",
      "purl": null
    },
    {
      "affected_notes": [],
      "affected_ranges": [],
      "ecosystem": null,
      "fixed_versions": [],
      "package_name": "rhoai/odh-ai-gateway-payload-processing-rhel9",
      "purl": null
    }
  ],
  "cwes": [
    639
  ],
  "cvss_score": 9.1,
  "cvss_src": "cna_vendor",
  "cvss_vector": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:L/A:L",
  "epss_score": null,
  "epss_percentile": null,
  "epss_asof": null,
  "automation": null,
  "check": false,
  "check_confidence_basis": null,
  "exploit": false,
  "pipeline_minutes": null,
  "t0_at": "2026-10-28T20:02:14.895000+00:00",
  "signals": {
    "trigger_signal": {
      "at": "2026-10-28T20:02:14.895000+00:00",
      "days": 0,
      "is_lower_bound": false,
      "measured_at": "2026-10-28T20:02:14.895000+00:00",
      "url": "https://www.cve.org/CVERecord?id=CVE-2026-117574"
    }
  }
}

published 2026-10-28 20:24Z
last change 2026-10-28 20:39Z

Need the technical details behind an entry?

Every entry here shows what we attempted and when. The technical details, the generated exploit that fired it and the evidence behind the verdict are not public. Tell us about your use case and we’ll get back to you about extended API access.