Summary
CVE-2026-42208 is a critical pre‑authentication SQL injection in the LiteLLM proxy (AI Gateway). The vulnerability exists in how the proxy constructs a database query used during API‑key verification: caller-supplied key data was incorporated directly into query text instead of being passed as a parameter. An attacker can reach this code path by sending a specially crafted Authorization header to any LLM API route (for example, POST /chat/completions) and trigger the proxy’s error handling to exercise the vulnerable query.
Affected software
- Product: LiteLLM (proxy / AI gateway)
- Versions: >= 1.81.16 and < 1.83.7
- Patched in: 1.83.7 (vendor recommends upgrading to the latest stable release)
Severity
- CVSS 3.1 Base Score: 9.8 (CRITICAL)
- Attack Vector: Network
- Privileges Required: None (pre‑auth)
- User Interaction: None
Impact
- Confidentiality: High — attackers can read sensitive data from the proxy database (including API keys and credentials managed by the proxy).
- Integrity: High — evidence indicates an attacker may be able to modify database records, potentially changing credentials or authorization state.
- Availability: High — database modification could also disrupt service.
Successful exploitation can lead to unauthorized access to the proxy, theft of managed credentials (API keys), and further compromise of dependent systems and services.
Observed activity and guidance
- The issue has been added to the CISA Known Exploited Vulnerabilities catalog and an action due date was published (CISA guidance recommending mitigations/patching).
- Security vendors and public reports indicated active exploitation shortly after disclosure; organizations should treat exposed LiteLLM proxies as high risk.
Mitigation and recommended actions
- Immediate action: Upgrade LiteLLM to version 1.83.7 or later (or the vendor-recommended latest stable release).
- If immediate upgrade is not possible:
- Restrict network access to the proxy (limit to trusted IPs / internal network only).
- Block or rate‑limit access to LLM API endpoints (e.g., POST /chat/completions) at perimeter controls or WAF.
- Rotate any API keys and credentials managed by the proxy after patching.
- Increase logging and monitoring for abnormal Authorization headers, unexpected DB errors, and unusual queries.
- Apply WAF rules to detect and block SQL injection patterns targeting API key parameters.
References to review before/after patching
- Review the vendor advisory and release notes for the exact patched versions and any additional mitigation guidance.
- After patching, validate integrity of the proxy database and rotate credentials if compromise is suspected.
IONIX Status
The IONIX research team is tracking ongoing exploitation attempts and recommends immediate patching. Potentially affected assets are outlined in this post.