LiteLLM SQL injection in proxy API key verification
Affects LiteLLM proxy versions 1.81.16 up to before 1.83.7, where the attack reaches the vulnerable query through error handling on any LLM API route. Fixed in 1.83.7.
- Exploited
- Yes, on CISA KEV
- Fix
- Fixed in 1.83.7
- Who is exposed
- Not reported
Summary
A SQL injection flaw in the LiteLLM proxy (BerriAI) lets an attacker read data from the proxy's database, and possibly modify it, which could give unauthorised access to the proxy and the credentials it manages. It can be reached remotely, without logging in, by sending a specially crafted Authorization header to any LLM API route, and public exploit code is available and attacks are happening.
Written by a model
Original description
LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. From version 1.81.16 to before version 1.83.7, a database query used during proxy API key checks mixed the caller-supplied key value into the query text instead of passing it as a separate parameter. An unauthenticated attacker could send a specially crafted Authorization header to any LLM API route (for example POST /chat/completions) and reach this query through the proxy's error-handling path. An attacker could read data from the proxy's database and may be able to modify it, leading to unauthorised access to the proxy and the credentials it manages. This issue has been patched in version 1.83.7.
Who should check
Teams running the LiteLLM proxy (AI Gateway, pip package litellm) versions 1.81.16 to before 1.83.7, especially any exposed to untrusted clients on LLM API routes.
“Affected versions >=1.81.16, <1.83.7”
What to do
Upgrade to a fixed version: 1.83.7.
Fix it by , CISA's deadline for US federal agencies.
CISA's required action: Apply mitigations per vendor instructions, follow applicable BOD 22-01 guidance for cloud services, or discontinue use of the product if mitigations are unavailable.
Check your systems with an agent
Paste this into a coding or ops agent. It checks your systems read-only and reports a table per asset.
Check whether my systems are affected by CVE-2026-42208 (LiteLLM: SQL injection in Proxy API key verification), using the cve.now data below.
Rules: work read-only. Do not run exploit code, proof-of-concept scripts or scanners that send attack payloads, and do not change any configuration, package, image or deployment. Report findings only.
What cve.now has for CVE-2026-42208 (entry last updated 2026-07-15):
Affected:
- BerriAI litellm: from 1.81.16, fixed in 1.83.7
Fixed versions: 1.83.7.
Conditions (all must be true for the bug to be hit):
- LiteLLM proxy running an affected version (1.81.16 up to before 1.83.7) (source: https://github.com/BerriAI/litellm/security/advisories/GHSA-r75f-5x8p-qvmc)
- The attack reaches the vulnerable query through the proxy's error-handling path on any LLM API route (source: https://github.com/BerriAI/litellm/security/advisories/GHSA-r75f-5x8p-qvmc)
Setup: unknown whether the bug needs a non-default feature or setting. Check the advisory.
Attacker needs: network access, no login, no user action.
CISA KEV: CVE-2026-42208 is listed as known exploited (added 2026-05-08, remediation due 2026-05-11). CISA's required action: Apply mitigations per vendor instructions, follow applicable BOD 22-01 guidance for cloud services, or discontinue use of the product if mitigations are unavailable.
This KEV status is for CVE-2026-42208 only. If you also check another CVE, look up its own KEV status; never assume it from this one.
How to check:
- Installed software: check what is actually running, not what a repo declares. Read the version from each host, container image and Kubernetes node (the binary's version output, the OS package database, image layers), and check whether the conditions above hold there. Distribution packages may carry a backported fix under an older version number; confirm against the distribution's own advisory.
Verify against the linked advisories before concluding, and cite the date you read each one:
- https://bugzilla.redhat.com/show_bug.cgi?id=2463965
- https://github.com/BerriAI/litellm/security/advisories/GHSA-r75f-5x8p-qvmc
- https://github.com/BerriAI/litellm/releases/tag/v1.83.7-stable
- https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42208
- https://access.redhat.com/security/cve/CVE-2026-42208
- https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-42208.json
Do not conclude an asset is unaffected unless you checked it and the evidence covers both its version and the conditions above. If you could not check something, mark it unknown and say what is missing.
Report a table with one row per asset and these columns: Asset | Version or build | Status (affected, fixed, not applicable, unknown) | Compensating controls | Exposure (internet-facing, internal, none) | Evidence | Next action.
Entry: https://cve.now/CVE-2026-42208
For follow-up questions, use the cve.now MCP server at https://mcp.cve.now/mcp (no authentication), for example execute with name cve.get and arguments {"id": "CVE-2026-42208"}.Affected versions
- BerriAI litellmFrom 1.81.16 · fixed in 1.83.7
cve.now Index
- Exploitation40/40Listed in CISA's Known Exploited Vulnerabilities catalog
- Exposure25/25Actively exploited, CISA rates it automatable
- Reach?/20Unknown. No tracked technology or package matched
- Severity10/10CVSS 9.3, total technical impact per CISA
- No patch0/5No signal
Impact
- Widely used“PyPI download statistics during March 2026 reflected approximately 3.4 million daily installations across direct users and downstream dependents” labs.cloudsecurityalliance.org62.3k stars on GitHub. berriai/litellm
Timeline
- CISA SSVC assessment
Exploitation: active · Automatable: yes · Technical impact: total
- CVE published
- Added to CISA KEV
CISA has evidence of exploitation in the wild
3 days - CISA remediation deadline
Due date for US federal agencies to patch
65 days - Record last updated123 days
- Exploit code on GitHub
wangp0tato/poc_cve-2026-42208
On X 2
Evidence 16
Questions about CVE-2026-42208
Is CVE-2026-42208 being exploited?
Yes. CISA added it to its Known Exploited Vulnerabilities catalog on 2026-05-08, which means it has evidence of attacks. US federal agencies must fix it by 2026-05-11. Public exploit code exists (Metasploit, a proof of concept on GitHub). Its EPSS score, the predicted chance of exploitation in the next 30 days, reached 10% or higher by 2026-10-27.
Source: cisa.gov, cve.now feed data
What does this mean?
The LiteLLM proxy builds a database query unsafely when it verifies API keys. An attacker can read data from the proxy's database and may be able to modify it. That could give unauthorised access to the proxy and the credentials it manages.
Source: github.com, security.access.redhat.com
Which versions are affected by CVE-2026-42208?
litellm: from 1.81.16 before 1.83.7.
Source: cve.now feed data
Am I affected, and how do I check?
LiteLLM proxy versions from 1.81.16 up to but not including 1.83.7 are affected. The vulnerable query is reached through the proxy's error-handling path on any LLM API route. To check a deployment, look at the installed litellm package version, for example with pip show litellm, or at the version of the proxy image in use.
Source: github.com, security.access.redhat.com
How do I fix CVE-2026-42208?
Upgrade LiteLLM to version 1.83.7 or later, which is the end of the affected range.
Source: github.com, security.access.redhat.com
Related
- CVE-2026-40217BerriAI LiteLLM Unprotected Alternate Channel Vulnerability
- CVE-2026-59822BerriAI LiteLLM Improper Authentication Vulnerability
- CVE-2026-42271BerriAI LiteLLM Command Injection Vulnerability
- CVE-2026-114612LiteLLM: Privilege Escalation to Proxy Admin via Cross-Domain Reuse of the Salt Key
- CVE-2026-49468LiteLLM: Authentication Bypass via Host Header Injection
- CVE-2026-89032BerriAI LiteLLM < 1.101.0-rc.1 Tenant Isolation Bypass via Semantic Cache Layer
No corrections so far. Something wrong on this page? Report an error.