/

CVE-2026-85178

Severity:

HIGH RISK

(7.7)

Broken Access Control (CWE-639) in Helicone Vault API

CVE-2026-85178

Deferred

Vulnerability type

Category:

Broken Access Control

CWE:

CWE-639

Impact:

Cross-tenant credential disclosure

Severity:

HIGH RISK

(7.7)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Helicone's vault key endpoint does not verify that the requesting administrator belongs to the organization associated with the requested provider key. As a result, an admin or owner from one organization may retrieve decrypted provider credentials belonging to another tenant.

Affected context

The vulnerability occurs in deployments where Helicone exposes the

/v1/vault/key/{providerKeyId} endpoint and permits organization-level

administrators or owners to request provider keys. Risk exists when the

endpoint relies on the supplied key identifier without enforcing an

organization-ownership relationship between the requester and the vault

record.

Exploit preconditions

  • Attack surface: The authenticated vault key retrieval API endpoint
  • Required access: Admin or owner privileges in any Helicone organization
  • Trigger point: A request for a provider key identified by

providerKeyId

  • Why exploitation is possible: The authorization path does not

consistently compare the requester's organization identifier with the

organization associated with the requested vault key

Why this happens

The underlying bug is an object-level authorization failure. The endpoint

retrieves and decrypts a provider credential based on an attacker-controlled

object identifier but does not enforce the required tenant ownership check

before returning the secret. This behavior maps to CWE-639: Authorization Bypass Through User-Controlled Key, where access to a resource is

determined by an identifier without adequately validating the caller's

authorization for that specific resource.

Potential impact

  • An attacker may disclose another organization's plaintext OpenAI,

Anthropic, or Bedrock credentials

  • Exposed provider keys could be used to incur unauthorized usage costs
  • Attackers could access or manipulate upstream AI services according to

the permissions of the disclosed credentials

  • Affected tenants could experience data exposure, service disruption, or

follow-on compromise

General mitigation guidance

Mitigation generally focuses on enforcing resource-level authorization

before retrieving or decrypting any secret.

Common defensive patterns include:

  • Deriving the requester organization from the authenticated session rather

than trusting request parameters

  • Enforcing an explicit organization-ownership check for every provider key

lookup

  • Applying deny-by-default authorization policies to vault resources
  • Separating authorization, retrieval, and decryption so unauthorized

records are never decrypted

  • Limiting administrative privileges and using narrowly scoped service

identities

  • Avoiding plaintext credential exposure when a reference, proxy, or

short-lived token can satisfy the integration requirement

How teams detect issues like this

Detection approaches can include:

  • Static analysis and code review rules requiring tenant or resource-owner

checks before object retrieval

  • API authorization tests that request valid resource identifiers across

organization boundaries

  • Review of handlers that accept IDs for secrets, billing objects, files, or

other tenant-owned resources

  • Audit-log monitoring for provider-key access where the requester and

resource organizations differ

  • Policy enforcement requiring object-level authorization decisions before

secret decryption or response serialization

Sources

SourceLink
github.comgithub.com/Helicone/helicone
github.comgithub.com/Helicone/helicone/commit/ca34549ea56f7ed587843f82d9cc19baa1f36ba4
github.comgithub.com/Helicone/helicone/issues/5712
vulncheck.comvulncheck.com/advisories/helicone-cross-tenant-provider-key-disclosure-via-missing-organization-filter

Frequently asked

questions

Yes, it may be exploitable through the authenticated vault key API when an

attacker has the required administrative role and can submit a provider key

identifier belonging to another organization.

The described condition requires authentication with admin or owner

privileges in an organization. It does not necessarily require membership

in the organization that owns the targeted key.

Enforce server-side, object-level authorization that compares the

authenticated organization with the provider key's organization before

retrieval, decryption, or response generation.

Use centralized tenant-aware authorization middleware, add cross-tenant

negative tests, minimize administrative scope, and monitor access to

sensitive resources and decrypted credentials.

Find out what's already
exploitable in your codebase.