Default Admin Key Exposed on 10% of Public LiteLLM Gateways Enables Full Access
What Happened — Research by Wiz in February found that roughly one‑in‑ten internet‑facing LiteLLM gateways accepted the example admin key “sk‑1234” that appears in the project’s own setup guide. The key grants unrestricted read/write access to the gateway, effectively handing an attacker full control over any connected AI models.
Why It Matters for Trust & Control Assurance
- This scenario is a textbook case of a missing “default credential” control, which continuous control‑assurance programs are built to detect, remediate, and evidence.
- Demonstrating that admin credentials are rotated and that configuration drift is monitored provides defensible audit evidence across multiple frameworks (e.g., NIST CSF, ISO 27001).
Who Is Affected — SaaS providers, AI‑focused tech firms, and any organization that deploys LiteLLM as an API gateway for model access.
Recommended Actions
- Enforce a policy that default credentials are disabled or rotated before any production deployment.
- Integrate automated configuration scans into your CI/CD pipeline to flag exposed admin keys and capture remediation evidence.
Technical Notes — The vulnerability stems from a misconfiguration: the LiteLLM gateway does not enforce credential change on first use, allowing the hard‑coded example key to authenticate. No CVE has been assigned; the issue is documented in Wiz’s public research. Source: The Hacker News