CVE-2026-62211: OpenClaw credential redaction

CISO Take

OpenClaw's trajectory export feature fails to properly redact credentials before handing data to lower-trust callers, letting an attacker with existing local, low-privilege access pull secrets that should stay inside the trusted boundary. There's no EPSS score, no CISA KEV listing, no public exploit code, and no Nuclei template for this one, and the CVSS vector requires local access, low complexity, low privileges, and user interaction — this is not a remotely exploitable, mass-scannable bug, but it's a real problem for any environment running OpenClaw agents with shared or multi-tenant access, since agent trajectories routinely capture API keys, tokens, and other secrets touched during a session. Blast radius is currently unverified since downstream dependent counts aren't available, but any organization using OpenClaw's trajectory export for debugging, auditing, or handoff workflows should treat exported trajectories as potentially containing live credentials until patched. Patch to OpenClaw 2026.6.1 or later, and in the interim restrict who can invoke the trajectory export feature and audit any already-exported trajectory files for embedded secrets, rotating anything found.

Sources: NVD GitHub Advisory ATLAS vulncheck.com

What is the risk?

Medium severity (CVSS 5.0, C:H/I:N/A:N) reflecting a pure confidentiality break with no integrity or availability impact. The attack requires local access, low privileges, and user interaction, which meaningfully narrows the exploitable population to insiders, compromised low-privilege accounts, or malicious co-tenants on shared OpenClaw deployments — not opportunistic internet-wide scanning. No EPSS data, no KEV listing, and no public PoC or scanner template are available, so near-term mass exploitation is unlikely. However, the impact is high-value: successful exploitation directly yields credentials, which can be chained into further compromise (API abuse, lateral movement, data exfiltration), so the real risk is concentrated in multi-user or CI/CD-adjacent OpenClaw deployments where trust boundaries between callers actually matter.

How does the attack unfold?

Local low-privilege access

Attacker obtains or already holds a low-privilege local account or service identity with access to the OpenClaw deployment.

Trigger trajectory export

Attacker invokes the trajectory export feature, exploiting a misconfigured input path or feature accessibility to bypass the trust boundary.

AML.T0055

Credential exposure

Redaction logic fails to strip sensitive data, so the exported trajectory includes credentials that should have remained trusted-boundary-only.

AML.T0083

Downstream abuse

Attacker uses the leaked credentials to access the same APIs, services, or tools the agent's trajectory had touched, escalating beyond the original low-privilege position.

AML.T0091

Local low-privilege access

Attacker obtains or already holds a low-privilege local account or service identity with access to the OpenClaw deployment.

Trigger trajectory export

Attacker invokes the trajectory export feature, exploiting a misconfigured input path or feature accessibility to bypass the trust boundary.

AML.T0055

Credential exposure

Redaction logic fails to strip sensitive data, so the exported trajectory includes credentials that should have remained trusted-boundary-only.

AML.T0083

Downstream abuse

Attacker uses the leaked credentials to access the same APIs, services, or tools the agent's trajectory had touched, escalating beyond the original low-privilege position.

AML.T0091

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
OpenClaw	pip	—	No patch
3 dependents 37% patched ~3d to patch Full package profile →

Do you use OpenClaw? You're affected.

How severe is it?

CVSS 3.1

5.0 / 10

EPSS

0.1%

chance of exploitation in 30 days

Higher than 5% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Trivial

What is the attack surface?

AV Local

AC Low

PR Low

UI Required

S Unchanged

C High

I None

A None

What should I do?

1 step

Upgrade OpenClaw to version 2026.6.1 or later, which fixes the redaction bypass. Until patched, restrict access to the trajectory export feature to trusted/high-privilege callers only, and review input path configuration for the export mechanism to close the misconfiguration the advisory references. Audit any trajectory exports already generated or shared for embedded credentials and rotate any secrets found. For detection, monitor for anomalous invocations of the export feature by low-privilege accounts or service identities that shouldn't normally use it, and log/alert on export requests that include unexpected input path parameters.

What does CISA's SSVC say?

Decision Track

Exploitation none

Automatable No

Technical Impact partial

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

How is it classified?

Data Leakage Auth Bypass Agent AML.T0037 - Data from Local System AML.T0055 - Unsecured Credentials AML.T0083 - Credentials from AI Agent Configuration

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15 - Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6 - AI system security controls

NIST AI RMF

MANAGE 4.1 - Risk monitoring and response for deployed AI systems

OWASP LLM Top 10

LLM02 - Sensitive Information Disclosure

Frequently Asked Questions

What is CVE-2026-62211?

OpenClaw's trajectory export feature fails to properly redact credentials before handing data to lower-trust callers, letting an attacker with existing local, low-privilege access pull secrets that should stay inside the trusted boundary. There's no EPSS score, no CISA KEV listing, no public exploit code, and no Nuclei template for this one, and the CVSS vector requires local access, low complexity, low privileges, and user interaction — this is not a remotely exploitable, mass-scannable bug, but it's a real problem for any environment running OpenClaw agents with shared or multi-tenant access, since agent trajectories routinely capture API keys, tokens, and other secrets touched during a session. Blast radius is currently unverified since downstream dependent counts aren't available, but any organization using OpenClaw's trajectory export for debugging, auditing, or handoff workflows should treat exported trajectories as potentially containing live credentials until patched. Patch to OpenClaw 2026.6.1 or later, and in the interim restrict who can invoke the trajectory export feature and audit any already-exported trajectory files for embedded secrets, rotating anything found.

Is CVE-2026-62211 actively exploited?

No confirmed active exploitation of CVE-2026-62211 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-62211?

Upgrade OpenClaw to version 2026.6.1 or later, which fixes the redaction bypass. Until patched, restrict access to the trajectory export feature to trusted/high-privilege callers only, and review input path configuration for the export mechanism to close the misconfiguration the advisory references. Audit any trajectory exports already generated or shared for embedded credentials and rotate any secrets found. For detection, monitor for anomalous invocations of the export feature by low-privilege accounts or service identities that shouldn't normally use it, and log/alert on export requests that include unexpected input path parameters.

What systems are affected by CVE-2026-62211?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, AI agent orchestration, credential/secrets handling in agent workflows.

What is the CVSS score for CVE-2026-62211?

CVE-2026-62211 has a CVSS v3.1 base score of 5.0 (MEDIUM). The EPSS exploitation probability is 0.15%.

What is the AI security impact?

Affected AI Architectures

agent frameworksAI agent orchestrationcredential/secrets handling in agent workflows

MITRE ATLAS Techniques

AML.T0037 Data from Local System

AML.T0055 Unsecured Credentials

AML.T0083 Credentials from AI Agent Configuration

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE 4.1

OWASP LLM Top 10: LLM02

What are the technical details?

Original Advisory

OpenClaw versions before 2026.6.1 contain a credential redaction bypass vulnerability in the trajectory export feature that allows lower-trust callers to access data that should remain within trusted boundaries. Attackers can exploit misconfigured input paths or feature accessibility to expose sensitive credentials and data through the export mechanism.

Exploitation Scenario

A lower-privilege user or service account with local access to an OpenClaw deployment — for example, a contractor account, a CI job, or a compromised low-tier credential — invokes the trajectory export feature, either directly or by manipulating an input path parameter the export mechanism exposes. Because the redaction logic doesn't correctly scope what should stay within the trusted boundary, the exported trajectory includes credentials (API keys, tokens) that were used during the agent's run rather than redacted. The attacker retrieves the export, extracts the leaked secrets, and uses them to escalate access — for instance, calling the same LLM API, cloud service, or internal tool the agent had access to, now under the compromised credential.

Weaknesses (CWE)

CWE-532 Insertion of Sensitive Information into Log File Primary CWE-532 Insertion of Sensitive Information into Log File

CWE-532 — Insertion of Sensitive Information into Log File: The product writes sensitive information to a log file.

[Architecture and Design, Implementation] Consider seriously the sensitivity of the information written into log files. Do not write secrets into the log files.
[Distribution] Remove debug log files before deploying the application into production.

Source: MITRE CWE corpus.