CVE-2026-32015: OpenClaw PATH hijack bypasses exec

CISO Take

CVE-2026-32015 is a CWE-426 (Untrusted Search Path) vulnerability in OpenClaw's `tools.exec.safeBins` allowlist that validates executable names but not their resolved filesystem paths, allowing a low-privileged attacker who controls the agent process PATH to substitute a trojan binary with an allowlisted name (e.g., `jq`) and achieve arbitrary code execution. AI agent frameworks like OpenClaw routinely run with elevated permissions and access sensitive data sources, meaning a successful exploit grants an attacker a highly privileged execution context within the AI pipeline. Although no public exploit exists and CISA KEV has not flagged this, the EPSS places this in the 97th percentile for exploitation likelihood relative to all tracked CVEs, and real-world abuse of OpenClaw's execution ecosystem has already been documented in AIID incident #1368. Organizations running OpenClaw versions prior to 2026.2.19 should patch immediately; where patching is delayed, harden agent launch environments by enforcing immutable PATH configurations and auditing process environment variables at gateway startup.

Sources: NVD EPSS GitHub Advisory ATLAS VulnCheck

What is the risk?

High risk for organizations deploying OpenClaw-based AI agents, particularly those granting agents tool-execution permissions. The attack requires only low privilege and no user interaction, making it trivially automatable once PATH influence is established. The EPSS 97th percentile indicates significantly elevated exploitation likelihood despite the low absolute probability. The local attack vector limits remote exploitation, but AI agent gateways are often accessible via CI/CD pipelines, developer workstations, and orchestration environments where PATH manipulation is feasible. The 396 other CVEs in the same package suggest a historically high-vulnerability codebase warranting elevated scrutiny.

How does the attack unfold?

Initial Access

Attacker with low-privilege access to the OpenClaw host places a trojan binary named after an allowlisted tool (e.g., `jq`) in a user-writable directory they control.

AML.T0074

PATH Manipulation

Attacker prepends their malicious directory to the OpenClaw gateway process PATH via shell profile injection, job definition tampering, or environment variable manipulation in the launch context.

AML.T0107

Allowlist Bypass

OpenClaw's `safeBins` validates the executable name 'jq' against the allowlist (check passes), but OS PATH resolution executes the trojan binary instead of the legitimate system binary.

AML.T0053

Code Execution & Impact

Trojan binary runs with the AI agent process's OS-level privileges, enabling credential exfiltration, reverse shell establishment, or lateral movement within the AI pipeline environment.

AML.T0112.000

Initial Access

Attacker with low-privilege access to the OpenClaw host places a trojan binary named after an allowlisted tool (e.g., `jq`) in a user-writable directory they control.

AML.T0074

PATH Manipulation

Attacker prepends their malicious directory to the OpenClaw gateway process PATH via shell profile injection, job definition tampering, or environment variable manipulation in the launch context.

AML.T0107

Allowlist Bypass

OpenClaw's `safeBins` validates the executable name 'jq' against the allowlist (check passes), but OS PATH resolution executes the trojan binary instead of the legitimate system binary.

AML.T0053

Code Execution & Impact

Trojan binary runs with the AI agent process's OS-level privileges, enabling credential exfiltration, reverse shell establishment, or lateral movement within the AI pipeline environment.

AML.T0112.000

What systems are affected?

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

Do you use OpenClaw? You're affected.

How severe is it?

CVSS 3.1

7.8 / 10

EPSS

0.1%

chance of exploitation in 30 days

Higher than 3% 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 None

S Unchanged

C High

I High

A High

What should I do?

5 steps

Patch immediately

Upgrade OpenClaw to 2026.2.19 or later (see commit 28bac46).
Workaround if patching is delayed

Enforce a fixed, immutable PATH for the OpenClaw gateway process at the process supervisor or systemd unit level; do not inherit PATH from user environments.
Detect

Audit process launch environment variables for OpenClaw processes; alert on PATH entries containing non-standard directories. Monitor for execution of allowlisted binaries (jq, curl, etc.) from unexpected filesystem paths using EDR process tree inspection.
Principle of least privilege

Run OpenClaw gateway processes in isolated containers with read-only filesystem mounts outside explicitly approved binary directories.
Supply chain hygiene

Verify checksums of allowlisted binaries at agent startup rather than relying solely on name-based validation.

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 Auth Bypass Supply Chain Agent Plugin Framework AML.T0050 - Command and Scripting Interpreter AML.T0053 - AI Agent Tool Invocation AML.T0074 - Masquerading AML.T0107 - Exploitation for Defense Evasion AML.T0112.000 - Local AI Agent

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15 - Accuracy, Robustness and Cybersecurity Article 9 - Risk Management System

ISO 42001

A.6.2 - AI Risk Assessment A.8.4 - AI System Security

NIST AI RMF

GOVERN 1.7 - Processes for AI Risk Management

OWASP LLM Top 10

LLM08 - Excessive Agency

Frequently Asked Questions

What is CVE-2026-32015?

CVE-2026-32015 is a CWE-426 (Untrusted Search Path) vulnerability in OpenClaw's `tools.exec.safeBins` allowlist that validates executable names but not their resolved filesystem paths, allowing a low-privileged attacker who controls the agent process PATH to substitute a trojan binary with an allowlisted name (e.g., `jq`) and achieve arbitrary code execution. AI agent frameworks like OpenClaw routinely run with elevated permissions and access sensitive data sources, meaning a successful exploit grants an attacker a highly privileged execution context within the AI pipeline. Although no public exploit exists and CISA KEV has not flagged this, the EPSS places this in the 97th percentile for exploitation likelihood relative to all tracked CVEs, and real-world abuse of OpenClaw's execution ecosystem has already been documented in AIID incident #1368. Organizations running OpenClaw versions prior to 2026.2.19 should patch immediately; where patching is delayed, harden agent launch environments by enforcing immutable PATH configurations and auditing process environment variables at gateway startup.

Is CVE-2026-32015 actively exploited?

No confirmed active exploitation of CVE-2026-32015 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-32015?

1. **Patch immediately**: Upgrade OpenClaw to 2026.2.19 or later (see commit 28bac46). 2. **Workaround if patching is delayed**: Enforce a fixed, immutable PATH for the OpenClaw gateway process at the process supervisor or systemd unit level; do not inherit PATH from user environments. 3. **Detect**: Audit process launch environment variables for OpenClaw processes; alert on PATH entries containing non-standard directories. Monitor for execution of allowlisted binaries (jq, curl, etc.) from unexpected filesystem paths using EDR process tree inspection. 4. **Principle of least privilege**: Run OpenClaw gateway processes in isolated containers with read-only filesystem mounts outside explicitly approved binary directories. 5. **Supply chain hygiene**: Verify checksums of allowlisted binaries at agent startup rather than relying solely on name-based validation.

What systems are affected by CVE-2026-32015?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, agentic pipelines, RAG pipelines, model serving.

What is the CVSS score for CVE-2026-32015?

CVE-2026-32015 has a CVSS v3.1 base score of 7.8 (HIGH). The EPSS exploitation probability is 0.13%.

What is the AI security impact?

Affected AI Architectures

agent frameworksagentic pipelinesRAG pipelinesmodel serving

MITRE ATLAS Techniques

AML.T0050 Command and Scripting Interpreter

AML.T0053 AI Agent Tool Invocation

AML.T0074 Masquerading

AML.T0107 Exploitation for Defense Evasion

AML.T0112.000 Local AI Agent

Compliance Controls Affected

EU AI Act: Article 15, Article 9

ISO 42001: A.6.2, A.8.4

NIST AI RMF: GOVERN 1.7

OWASP LLM Top 10: LLM08

What are the technical details?

Original Advisory

OpenClaw versions 2026.1.21 prior to 2026.2.19 contain a path hijacking vulnerability in tools.exec.safeBins that allows attackers to bypass allowlist checks by controlling process PATH resolution. Attackers who can influence the gateway process PATH or launch environment can execute trojan binaries with allowlisted names, such as jq, circumventing executable validation controls.

Exploitation Scenario

An attacker with low-privilege access to the host running an OpenClaw AI agent gateway (e.g., via a compromised developer workstation or misconfigured CI runner) creates a malicious binary named `jq` in a user-writable directory such as `~/.local/bin`. They then modify the gateway process's PATH—either via shell profile injection, environment variable manipulation in a job definition, or a malicious wrapper script—so that `~/.local/bin` precedes `/usr/bin`. When OpenClaw next invokes `jq` as part of its normal agent tool pipeline (e.g., parsing API responses), `safeBins` validates the name 'jq' against the allowlist (passing), but resolves execution to the trojan binary. The trojan executes with the agent's OS privileges, establishing a reverse shell or exfiltrating cloud credentials, model artifacts, or secrets stored in the agent's environment.

Weaknesses (CWE)

CWE-426 Untrusted Search Path

CWE-426 — Untrusted Search Path: The product searches for critical resources using an externally-supplied search path that can point to resources that are not under the product's direct control.

[Architecture and Design, Implementation] Hard-code the search path to a set of known-safe values (such as system directories), or only allow them to be specified by the administrator in a configuration file. Do not allow these settings to be modified by an external party. Be careful to avoid related weaknesses such as CWE-426 and CWE-428.
[Implementation] When invoking other programs, specify those programs using fully-qualified pathnames. While this is an effective approach, code that uses fully-qualified pathnames might not be portable to other systems that do not use the same pathnames. The portability can be improved by locating the full-qualified paths in a centralized, easily-modifiable location within the source code, and having the code refer to these paths.

Source: MITRE CWE corpus.