Severity by source
CVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
Primary rating from NVD · only source for this CVE.
CVSS VectorNVD
CVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
Lifecycle Timeline
5Blast Radius
ecosystem impact- 1 npm packages depend on openclaw (1 direct, 0 indirect)
Ecosystem-wide dependent count for version 2026.4.23.
DescriptionCVE.org
OpenClaw before 2026.4.23 contains an improper access control vulnerability in the gateway tool's config.apply and config.patch operations that allows compromised models to write unsafe configuration changes by bypassing an incomplete denylist protection. Attackers can persist malicious config modifications affecting command execution, network behavior, credentials, and operator policies that survive restart.
AnalysisAI
Compromised AI models running with access to OpenClaw's gateway tool can persist malicious configuration changes affecting command execution, network endpoints, credentials, and security policies by exploiting an incomplete denylist that failed to protect newly added config paths. The vulnerability allows model-driven writes to sensitive config subtrees (command execution safeguards, proxy/TLS settings, telemetry hooks, operator policies) that survive restart, enabling persistent control beyond the intended model-to-operator trust boundary. Patch available in OpenClaw 2026.4.23 with fail-closed allowlist enforcement. No public exploit or active exploitation confirmed, but CVSS 8.8 (AV:N/AC:L/PR:L) indicates authenticated remote attackers with low-privilege model access can achieve full config compromise.
Technical ContextAI
OpenClaw is an AI agent framework (npm package) that provides models with a 'gateway' tool for runtime configuration mutation. The vulnerability exists in the config.apply and config.patch operations' access control guard, which relied on a hand-maintained denylist of protected config paths. As OpenClaw's config schema expanded, new sensitive paths (tools.exec.safeBins, gateway.remote.url, hooks endpoints, credential forwarding settings, memory/indexing surfaces, operator policy controls) were not added to the denylist, creating exploitable gaps. This is a CWE-184 (Incomplete List of Disallowed Inputs) vulnerability where the security boundary depends on blocking known-bad inputs rather than allowing known-good ones. The affected component is the agent-facing gateway tool RPC layer in the npm package cpe:2.3:a:openclaw:openclaw:*:*:*:*:*:*:*:* before version 2026.4.23. The fix replaces the denylist with a fail-closed allowlist permitting only narrow agent-tunable paths (prompt/model settings, mention-gating configs like Telegram requireMention) and rejecting all other config mutations before the RPC is invoked.
RemediationAI
Upgrade to OpenClaw 2026.4.23 or later, available via npm (npm install openclaw@2026.4.23 or npm install openclaw@latest). The fix commit bceda6089aa7b3695cc7696b43c61ae3d01bb0ec replaced the incomplete denylist with a fail-closed allowlist that permits only narrow agent-tunable paths (prompt/model settings, mention-gating configs) and rejects all other config mutations before gateway RPC invocation. If immediate upgrade is not possible, compensating controls include: (1) Disable or restrict gateway tool access for models processing untrusted input by removing 'gateway' from the tools.allow list in operator config - this eliminates the attack surface but also disables legitimate model-driven config tuning; (2) Run OpenClaw in read-only config mode if the deployment supports it, preventing all runtime config writes - this breaks dynamic model customization workflows; (3) Deploy models in isolated containers/VMs with separate config stores and restart policies that revert to known-good config on reboot - this limits persistence but does not prevent in-session exploitation. Vendor advisory at https://github.com/openclaw/openclaw/security/advisories/GHSA-cwj3-vqpp-pmxr and patch commit at https://github.com/openclaw/openclaw/commit/bceda6089aa7b3695cc7696b43c61ae3d01bb0ec provide additional technical detail.
Auth bypass in OpenClaw voice-call extension before 2026.2.1. EPSS 0.68%. PoC and patch available.
Privilege escalation in OpenClaw (pre-2026.3.28) allows unauthenticated remote attackers to gain administrative access b
OpenClaw versions 2026.2.22 through 2026.2.24 contain a privilege escalation vulnerability that allows authenticated att
An authorization mismatch vulnerability in OpenClaw versions prior to 2026.3.1 allows authenticated users with operator.
OpenClaw versions prior to 2026.1.29 automatically establish WebSocket connections to attacker-controlled gateway URLs e
Path traversal in OpenClaw through version 2026.3.23 enables unauthenticated remote attackers to read arbitrary files in
OpenClaw sandbox browser functionality launches x11vnc for noVNC observer sessions without requiring authentication, all
OpenClaw versions before 2026.2.26 allow authenticated attackers to write arbitrary files outside the workspace director
OpenClaw versions prior to 2026.2.22 contain a shell environment variable injection vulnerability in the system.run func
OpenClaw versions prior to 2026.2.22 contain a resource exhaustion vulnerability where the application fails to consiste
OpenClaw versions prior to 2026.3.1 contain a sandbox escape vulnerability that allows authenticated attackers with low
OpenClaw versions 2026.1.30 and below fail to validate Telegram webhook secret tokens when `channels.telegram.webhookSec
Same weakness CWE-184 – Incomplete List of Disallowed Inputs
View allSame technique Authentication Bypass
View allShare
External POC / Exploit Code
Leaving vuln.today
EUVD-2026-29151
GHSA-8wcm-622f-3r46