
Cloud Vulnerability DB
A community-led vulnerabilities database
CVE-2026-27486 is an unverified process ownership vulnerability (CWE-283) in the OpenClaw CLI, a personal AI assistant. Versions 2026.2.13 and below are affected; the issue was disclosed on February 16, 2026, and patched in version 2026.2.14 released February 15, 2026. The vulnerability allows the CLI's process cleanup mechanism to terminate unrelated processes on shared hosts by matching command-line patterns without verifying process ownership. It carries a CVSS v3.1 base score of 5.3 (Medium) (GitHub Advisory, Red Hat CVE).
The root cause is classified as CWE-283 (Unverified Ownership): the CLI runner cleanup helpers (cleanupSuspendedCliProcesses and cleanupResumeProcesses) used pkill -f <pattern> or system-wide ps enumeration with regex pattern matching to identify and kill processes, without verifying that matched processes were children of the current OpenClaw process (i.e., without checking ppid == process.pid). This means any process on the same host whose command-line arguments matched the pattern could be sent SIGKILL. The fix scopes termination to direct child PIDs by filtering ps output on the parent PID field, adds graceful SIGTERM before SIGKILL, and tightens token matching to avoid substring false positives (GitHub Commit 6084d13, GitHub Commit eb60e2e, GitHub Advisory).
The primary impact is availability: on shared or multi-tenant hosts, the OpenClaw cleanup mechanism could inadvertently terminate unrelated processes belonging to other users or services if their command-line arguments matched the cleanup pattern. This could cause unexpected service interruptions, application crashes, or system instability for co-located workloads. There is no confidentiality or integrity impact; the vulnerability does not enable data access or modification (GitHub Advisory, Red Hat CVE).
There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the time of disclosure (Red Hat CVE). The EPSS score is approximately 0.013% (0.000130), indicating very low probability of exploitation in the near term. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog. Exploitation would require the attacker to have a process running on the same shared host with a command-line matching OpenClaw's cleanup patterns, and would depend on OpenClaw's cleanup routine being triggered (GitHub Advisory).
codex exec resume <sessionId> or claude --session-id <id>).ps -ax and applies pattern matching./var/log/syslog or journalctl) showing processes killed with SIGKILL without a clear initiating user action.codex, resume, claude, or --session-id on shared hosts running OpenClaw CLI ≤ 2026.2.13.Upgrade OpenClaw CLI to version 2026.2.14 or later, which scopes process cleanup to direct child PIDs only (filtering by ppid == process.pid) and adds graceful SIGTERM before SIGKILL (GitHub Release, GitHub Advisory). As a workaround prior to patching, deploy OpenClaw CLI in isolated environments or dedicated hosts rather than shared multi-tenant systems to eliminate exposure. Monitor process termination events and system logs for unexpected service interruptions that may indicate the cleanup mechanism is affecting unrelated processes.
The vulnerability was reported by @aether-ai-agent and credited in the official GitHub Security Advisory published by the OpenClaw maintainer (steipete) (GitHub Advisory). A technical write-up was published at infinitsec.net shortly after disclosure, describing the issue as "Unvalidated PID Kill via SIGKILL in Process Cleanup" (infinitsec.net). No significant broader media coverage or notable community controversy has been observed, consistent with the vulnerability's low severity rating.
Source: This report was generated using AI
Free Vulnerability Assessment
Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.
Get a personalized demo
"Best User Experience I have ever seen, provides full visibility to cloud workloads."
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
"We know that if Wiz identifies something as critical, it actually is."