Skip to main content
Vulnerability Database/CVE-2026-100575

CVE-2026-100575: OpenClaw Slack Auth Bypass Vulnerability

CVE-2026-100575 is an authentication bypass flaw in OpenClaw Slack that allows unauthorized users to trigger agents in group messages. This post explains its impact, affected versions, and mitigation steps.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2026-100575 Overview

CVE-2026-100575 is a missing authorization vulnerability [CWE-862] in OpenClaw Slack versions before 2026.8.1. The flaw allows disallowed participants in multi-person direct messages (group DMs) to invoke Slack agents and reach tools and data that those agents can access. The root cause is improper enforcement of sender allowlists configured for agent interactions.

An attacker with low-privileged Slack access can bypass configured sender policies by routing messages through a group DM that includes an allowed participant. This grants unauthorized use of connected agent tools and exposure to data those agents are permitted to retrieve.

Critical Impact

Low-privileged Slack users can bypass agent sender allowlists via group DMs, invoking agent tooling and accessing data outside their authorization boundary.

Affected Products

  • OpenClaw Slack versions prior to 2026.8.1
  • Deployments relying on sender allowlists for agent access control
  • Multi-person direct message (group DM) workflows integrated with OpenClaw agents

Discovery Timeline

  • 2026-09-26 - CVE-2026-100575 published to the National Vulnerability Database (NVD)
  • 2026-09-28 - Last updated in NVD database

Technical Details for CVE-2026-100575

Vulnerability Analysis

OpenClaw Slack provides agent integrations that execute tool calls and retrieve data on behalf of Slack users. Administrators restrict which users may invoke each agent by configuring a sender allowlist. The vulnerability stems from the allowlist check running against an incomplete view of participants in multi-person direct messages.

When a message arrives through a group DM, the authorization logic does not evaluate every sender in the conversation. A disallowed participant can issue agent-triggering content inside the shared channel, and the agent processes the request as if it came from an authorized party. The agent then returns tool output or data that the disallowed participant should not receive.

The impact spans confidentiality, integrity, and availability of the data and tools exposed to the agent. Any secrets, connectors, or write-capable integrations granted to the agent become reachable to users outside the allowlist.

Root Cause

The underlying weakness is a missing authorization check [CWE-862]. The sender policy evaluation does not iterate across all participants of a group DM before permitting agent invocation. The allowlist is enforced at the conversation level rather than at the per-sender level.

Attack Vector

Exploitation requires network access to the Slack workspace and low-privileged user credentials. No user interaction from the targeted agent operator or administrator is needed. The attacker joins or is invited to a group DM that also contains an allowed sender and then transmits agent-directed content.

See the GitHub Security Advisory GHSA-58qx-6m8p-wh2j and the VulnCheck Advisory - OpenClaw Authentication Bypass for additional technical context.

Detect Identity Attacks the Moment Authentication Fails

Singularity™ Identity surfaces credential misuse, lateral movement, and privilege escalation in real time, so you can contain the breach before it spreads.

Stop Identity Attacks

Detection Methods for CVE-2026-100575

Indicators of Compromise

  • Agent invocations originating from group DMs where the triggering participant is not on the configured sender allowlist
  • Unexpected tool-call activity in OpenClaw agent logs tied to multi-person conversation identifiers
  • Data retrieval or outbound integration calls initiated by agents outside normal business hours or baseline patterns

Detection Strategies

  • Correlate OpenClaw agent audit logs with Slack conversation metadata to identify sender identities for every agent trigger
  • Alert on any agent invocation where a participant in the source conversation is absent from the agent's sender allowlist
  • Review tool-call history for data access patterns that cross established authorization boundaries

Monitoring Recommendations

  • Ingest OpenClaw and Slack audit events into a centralized logging platform with retention aligned to incident response needs
  • Build detections that enumerate all participants of a group DM at the time of agent invocation, not just the direct sender
  • Track changes to agent sender allowlist configuration and flag silent policy drift

How to Mitigate CVE-2026-100575

Immediate Actions Required

  • Upgrade OpenClaw Slack to version 2026.8.1 or later across all deployed instances
  • Audit existing group DMs that interact with OpenClaw agents and review historical tool-call activity for unauthorized use
  • Rotate credentials, tokens, and API keys exposed to any agent that may have serviced unauthorized requests

Patch Information

OpenClaw addressed the authorization gap in version 2026.8.1. Patched builds enforce the sender allowlist against every participant in a multi-person direct message before an agent executes. Refer to the GitHub Security Advisory GHSA-58qx-6m8p-wh2j for release details.

Workarounds

  • Disable OpenClaw agents in multi-person direct message contexts until the upgrade is deployed
  • Restrict agents to single-user DMs or named channels where participant lists can be governed centrally
  • Remove high-impact tools and data connectors from agents that remain reachable from shared conversations
bash
# Configuration example
# Verify the installed OpenClaw Slack version
openclaw --version

# Upgrade to the patched release
pip install --upgrade "openclaw-slack>=2026.8.1"

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.