OpenClaw — On Our Radar

Status: 🟡 Under observation
Why we're talking about it: The framework has spiked to 180,000 GitHub stars following news that creator Peter Steinberger joined OpenAI to lead their agent strategy (Source: Sam Altman on X).

What we know:
- OpenClaw orchestrates frontier models including GPT-5 and Claude 4.5 Opus to manage local system tasks and digital operations (Source: haimaker.ai).
- Critical vulnerabilities CVE-2026-25253 and CVE-2026-25157 permit 1-click Remote Code Execution and OS command injection (Source: Cyera Labs).
- The architecture is currently a mass of technical debt, with 6,600+ "vibe-coded" commits in January 2026 largely generated by AI without manual audit (Source: cubic.dev).
- The project is transitioning to an independent foundation supported by OpenAI as of February 2026 (Source: Benzinga).
- Over 135,000 instances were exposed in early 2026 due to the tool's reliance on broad system-level permissions (Source: NSFocus).

The unknowns:
- We don't know yet if a formal security audit has been conducted on version 2026.2.x following the OpenAI hiring announcement.
- Clarity is missing regarding how OpenAI's involvement will affect the project's model-agnostic nature.

This article will be updated when we have more data. Until then, proceed with caution. Relying on 6,600 unaudited AI commits is a bold strategy for anyone who values their root directory.

Ship clean code,
Marcus.