Severity by source
AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N
Network-reachable via Kibana API, low complexity once realm overlap exists, low-privilege authenticated session required, high confidentiality impact on agent config, low integrity impact from modification or deletion.
Primary rating from Vendor (elastic).
CVSS VectorVendor: elastic
Lifecycle Timeline
2DescriptionCVE.org
Kibana Agent Builder determines whether a caller owns a private agent by comparing a stable user identifier when one is recorded, and falling back to a comparison of the username when it is not. A username is not unique across Elasticsearch authentication realms, so two distinct principals that share a username in different realms are treated as the same owner. This discloses the configuration and instructions of an agent the caller does not own, and allows that agent to be altered or removed.
AnalysisAI
Incorrect authorization in Kibana Agent Builder allows authenticated users to access, modify, or delete private agents they do not own when sharing a username with another principal across different Elasticsearch authentication realms. The flaw stems from a fallback ownership check that compares usernames - which are not unique across realms - rather than stable user identifiers, enabling cross-realm identity confusion. No public exploit or active exploitation (CISA KEV) has been identified at time of analysis, but the low attack complexity and network accessibility make this a credible insider or multi-tenant threat.
Technical ContextAI
Kibana Agent Builder is a feature within Elastic's Kibana UI (CPE: cpe:2.3:a:elastic:kibana:*:*:*:*:*:*:*:*) that allows users to create and manage private AI agents with custom configurations and instructions. Elasticsearch supports multiple authentication realms simultaneously - for example, a native realm alongside an LDAP or Active Directory realm - and a 'username' string is only unique within a single realm, not globally across the stack. The Agent Builder's ownership check uses a stable user identifier (such as a realm-scoped principal UID) as the primary comparator, but silently falls back to a plain username string comparison when no stable identifier is recorded. This fallback is the vulnerable code path: CWE-863 (Incorrect Authorization) describes exactly this pattern, where an authorization decision is made using an insufficiently precise principal identity. An attacker in Realm A whose username matches a victim in Realm B can exploit the fallback path to be incorrectly treated as the same owner.
RemediationAI
Upgrade Kibana to version 9.4.5 or 9.5.1 per Elastic security advisory ESA-2026-124 (https://discuss.elastic.co/t/kibana-9-4-5-9-5-1-security-update-esa-2026-124/389538). These are the patch versions identified in the advisory title; no earlier patched minor release has been confirmed. If immediate upgrade is not possible, a targeted compensating control is to audit and eliminate username collisions across all configured Elasticsearch authentication realms - ensuring no two principals in different realms share an identical username string removes the condition that triggers the fallback code path. This control has a trade-off: it may require coordinated username changes with identity providers (LDAP/AD) and could disrupt existing user mappings. A second workaround is to restrict Agent Builder usage to a single authentication realm by temporarily disabling multi-realm configurations, though this may impact SSO or identity federation workflows. Do not rely on access controls at the network perimeter alone, as any authenticated Kibana user in a multi-realm deployment can trigger this flaw.
Kibana versions before 5.6.15 and 6.6.1 contain an arbitrary code execution flaw in the Timelion visualizer. Rated criti
Kibana versions 6.7.0 to 6.8.8 and 7.0.0 to 7.6.2 contain a prototype pollution flaw in the Upgrade Assistant. Rated hig
Kibana versions before 6.4.3 and 5.6.13 contain an arbitrary file inclusion flaw in the Console plugin. Rated critical s
Prototype pollution in Kibana leads to arbitrary code execution via a crafted file upload and specifically crafted HTTP
Kibana versions 4.0 to 4.6, 5.0 to 5.6.12, and 6.0 to 6.4.2 contain an error in the way authorization credentials are us
A Prototype pollution vulnerability in Kibana leads to arbitrary code execution via crafted HTTP requests to machine lea
Kibana versions before 6.6.1 contain an arbitrary code execution flaw in the security audit logger. Rated critical sever
Prototype Pollution in Kibana can lead to code injection via unrestricted file upload combined with path traversal. Rate
Privilege escalation in Elastic Kibana's Agent Builder component (versions 9.4.0-9.4.3) allows authenticated low-privile
A deserialization issue in Kibana can lead to arbitrary code execution when Kibana attempts to parse a YAML document con
Kibana version 8.7.0 contains an arbitrary code execution flaw. Rated high severity (CVSS 8.8), this vulnerability is re
Kibana versions 8.0.0 through 8.7.0 contain an arbitrary code execution flaw. Rated high severity (CVSS 8.8), this vulne
Same weakness CWE-863 – Incorrect Authorization
View allSame technique Authentication Bypass
View allShare
External POC / Exploit Code
Leaving vuln.today
EUVD-2026-58288
GHSA-m5pm-2p35-xvpr