CVE-2026-42208: BerriAI LiteLLM SQL Injection Vulnerability

BerriAI LiteLLM contains a SQL injection vulnerability tracked as CVE-2026-42208. The flaw allows an attacker to read data from the proxy's database and potentially modify it, which can result in unauthorized access to the proxy itself and to the credentials it manages.

How it works

The weakness is categorized as CWE-89, SQL injection. An attacker supplies crafted input that alters the structure of database queries executed by the LiteLLM proxy. Successful exploitation can expose stored data or permit changes to records, directly affecting the proxy's configuration and any credentials held in its database.

Am I affected? How to find it in your systems

Inventory all deployments of BerriAI LiteLLM, including containerized instances, cloud-hosted proxies, and any environments where it fronts LLM API calls. Confirm the presence and configuration of the database component referenced in the CISA summary. Review database query logs and proxy access logs for anomalous patterns such as unexpected input strings or queries that deviate from expected application behavior. Specific version and configuration details must be checked against the vendor advisory.

How to remediate

Apply mitigations per the vendor instructions. Follow applicable BOD 22-01 guidance for cloud services. Where mitigations are unavailable, discontinue use of the product. After applying updates, validate that the database interaction paths no longer accept unsanitized input from untrusted sources.

If you can't patch immediately

Restrict network access to the LiteLLM proxy so that only trusted clients can reach it. Monitor database and proxy logs for signs of unexpected query structure or data access. Consider disabling features that expose the affected database paths until remediation is complete. If the instance runs in a cloud environment, apply the relevant BOD 22-01 controls.

If your data may have been exposed

Actively exploited vulnerabilities of this class have led to breaches involving credential and configuration data. Organizations can run a free exposure scan of their email domains to check for presence in known breach data sets.