CVE-2026-62217: OpenClaw QQBot exec approval bypass

CISO Take

OpenClaw's QQBot exec-approvals feature — built to gate which senders are trusted to trigger command execution through the agent — has an authorization flaw (CWE-863, incorrect authorization) that lets a lower-trust or non-allowlisted caller execute or persist actions beyond their intended permissions whenever the feature is enabled and network-reachable. For an AI agent framework whose entire value proposition is executing actions on a user's behalf, this is effectively unauthorized remote command execution achieved through a broken permission check rather than a memory-safety bug, and the CVSS 8.8 score (AV:N/AC:L/PR:L/UI:N, full C/I/A impact) reflects that it needs no user interaction and only low attacker privileges. There's no EPSS score, no CISA KEV listing, and no public exploit or Nuclei template yet, so mass opportunistic exploitation hasn't been observed, but both the OpenClaw maintainers and VulnCheck have already published advisories (GHSA-7jx6-764p-fgg9), which tends to draw researcher attention quickly. Any deployment running OpenClaw 2026.5.14-beta.1 through before 2026.5.27 with QQBot exec approvals enabled should upgrade to 2026.5.27 immediately; if immediate patching isn't possible, disable or network-isolate the exec-approvals endpoint and audit agent action logs for exec/persist events originating from senders outside the configured allowlist.

Sources: NVD GitHub Advisory CISA KEV ATLAS

What is the risk?

Technically trivial to exploit once reachable: no user interaction, low attacker privileges, low attack complexity, and network access are all that's required (CVSS 8.8). The gating factor is exposure — the flaw only matters where the QQBot exec approvals feature is enabled and reachable from a lower-trust network or chat surface. Where that holds, impact is severe: full confidentiality, integrity, and availability loss, since the vulnerable component is the very control meant to gate command execution by the agent. Absence from CISA KEV, no EPSS score, and no public exploit/scanner template lower near-term mass-exploitation risk, but the published GHSA and VulnCheck advisories mean technical detail is already public, and authorization-bypass bugs in agent exec paths are attractive, easily weaponized targets.

How does the attack unfold?

Unauthorized Reach

A non-allowlisted, lower-trust sender reaches the QQBot exec-approvals interface, which is network-accessible when the feature is enabled.

AML.T0049

Authorization Bypass

The exec-approvals authorization check fails to correctly enforce the caller's trust level, letting an unapproved request proceed as if authorized.

AML.T0107

Unauthorized Execution

The AI agent executes or persists an action requested by the unauthorized caller, exceeding the caller's intended permissions.

AML.T0053

Impact

The attacker achieves confidentiality, integrity, and availability compromise of the host or environment running the OpenClaw agent.

Unauthorized Reach

A non-allowlisted, lower-trust sender reaches the QQBot exec-approvals interface, which is network-accessible when the feature is enabled.

AML.T0049

Authorization Bypass

The exec-approvals authorization check fails to correctly enforce the caller's trust level, letting an unapproved request proceed as if authorized.

AML.T0107

Unauthorized Execution

The AI agent executes or persists an action requested by the unauthorized caller, exceeding the caller's intended permissions.

AML.T0053

Impact

The attacker achieves confidentiality, integrity, and availability compromise of the host or environment running the OpenClaw agent.

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

8.8 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 38% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Moderate

What is the attack surface?

AV Network

AC Low

PR Low

UI None

S Unchanged

C High

I High

A High

What should I do?

1 step

1) Upgrade OpenClaw to 2026.5.27 or later immediately — this is the vendor-fixed version per the GHSA advisory. 2) If upgrading isn't immediately possible, disable the QQBot exec approvals feature or restrict its network reachability (firewall/VPN-only access) until patched. 3) Review and tighten the sender allowlist configuration to confirm it reflects intended trust boundaries. 4) Audit agent/exec logs for command execution or persistence events triggered by senders not on the configured allowlist, especially during the 2026.5.14-beta.1 → 2026.5.27 exposure window. 5) Treat any exec-approval bypass indicators as a potential compromise and rotate credentials/tokens accessible to the affected OpenClaw instance.

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?

Auth Bypass Code Execution Agent Plugin AML.T0049 - Exploit Public-Facing Application AML.T0053 - AI Agent Tool Invocation AML.T0107 - Exploitation for Defense Evasion

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 operation and monitoring

NIST AI RMF

MEASURE 2.7 - AI system security and resilience are evaluated and documented

OWASP LLM Top 10

LLM06:2025 - Excessive Agency

Frequently Asked Questions

What is CVE-2026-62217?

OpenClaw's QQBot exec-approvals feature — built to gate which senders are trusted to trigger command execution through the agent — has an authorization flaw (CWE-863, incorrect authorization) that lets a lower-trust or non-allowlisted caller execute or persist actions beyond their intended permissions whenever the feature is enabled and network-reachable. For an AI agent framework whose entire value proposition is executing actions on a user's behalf, this is effectively unauthorized remote command execution achieved through a broken permission check rather than a memory-safety bug, and the CVSS 8.8 score (AV:N/AC:L/PR:L/UI:N, full C/I/A impact) reflects that it needs no user interaction and only low attacker privileges. There's no EPSS score, no CISA KEV listing, and no public exploit or Nuclei template yet, so mass opportunistic exploitation hasn't been observed, but both the OpenClaw maintainers and VulnCheck have already published advisories (GHSA-7jx6-764p-fgg9), which tends to draw researcher attention quickly. Any deployment running OpenClaw 2026.5.14-beta.1 through before 2026.5.27 with QQBot exec approvals enabled should upgrade to 2026.5.27 immediately; if immediate patching isn't possible, disable or network-isolate the exec-approvals endpoint and audit agent action logs for exec/persist events originating from senders outside the configured allowlist.

Is CVE-2026-62217 actively exploited?

No confirmed active exploitation of CVE-2026-62217 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-62217?

1) Upgrade OpenClaw to 2026.5.27 or later immediately — this is the vendor-fixed version per the GHSA advisory. 2) If upgrading isn't immediately possible, disable the QQBot exec approvals feature or restrict its network reachability (firewall/VPN-only access) until patched. 3) Review and tighten the sender allowlist configuration to confirm it reflects intended trust boundaries. 4) Audit agent/exec logs for command execution or persistence events triggered by senders not on the configured allowlist, especially during the 2026.5.14-beta.1 → 2026.5.27 exposure window. 5) Treat any exec-approval bypass indicators as a potential compromise and rotate credentials/tokens accessible to the affected OpenClaw instance.

What systems are affected by CVE-2026-62217?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, AI agent tool invocation pipelines, chat-integrated command execution workflows.

What is the CVSS score for CVE-2026-62217?

CVE-2026-62217 has a CVSS v3.1 base score of 8.8 (HIGH). The EPSS exploitation probability is 0.45%.

What is the AI security impact?

Affected AI Architectures

agent frameworksAI agent tool invocation pipelineschat-integrated command execution workflows

MITRE ATLAS Techniques

AML.T0049 Exploit Public-Facing Application

AML.T0053 AI Agent Tool Invocation

AML.T0107 Exploitation for Defense Evasion

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.6

NIST AI RMF: MEASURE 2.7

OWASP LLM Top 10: LLM06:2025

What are the technical details?

Original Advisory

OpenClaw 2026.5.14-beta.1 before 2026.5.27 contain an authorization flaw in the QQBot exec approvals feature. When the feature is enabled and reachable, a lower-trust caller or configured input path could execute or persist actions beyond the caller's intended authorization, allowing non-allowlisted senders to perform unauthorized operations.

Exploitation Scenario

An attacker who is not on OpenClaw's configured allowlist — but who can reach the QQBot integration, e.g., as a low-privilege chat participant or via a spoofed/misrouted input path — sends a request that should require allowlisted approval to trigger command execution. Because the authorization check in the exec-approvals feature fails to properly validate the caller's trust level, the request is processed as if it came from an approved sender, causing the agent to execute or persist an arbitrary action (such as a shell command or file operation) on the host running OpenClaw. This gives the attacker a foothold equivalent to unauthorized remote code execution, without needing valid credentials beyond low-level network/chat access.

Weaknesses (CWE)

CWE-863 Incorrect Authorization Primary CWE-863 Incorrect Authorization

CWE-863 — Incorrect Authorization: The product performs an authorization check when an actor attempts to access a resource or perform an action, but it does not correctly perform the check.

[Architecture and Design] Divide the product into anonymous, normal, privileged, and administrative areas. Reduce the attack surface by carefully mapping roles with data and functionality. Use role-based access control (RBAC) [REF-229] to enforce the roles at the appropriate boundaries. Note that this approach may not protect against horizontal authorization, i.e., it will not protect a user from attacking others with the same role.
[Architecture and Design] Ensure that access control checks are performed related to the business logic. These checks may be different than the access control checks that are applied to more generic resources such as files, connections, processes, memory, and database records. For example, a database may restrict access for medical records to a specific database user, but each record might only be intended to be accessible to the patient and the patient's doctor [REF-7].

Source: MITRE CWE corpus.