Find notable cyber news and cases, enriched with sources, timelines, and signals.
Home Browse News Feed Stats Watchlists Beta About Contact

LiteLLM pre-auth SQL injection (CVE-2026-42208)

Vulnerability
First reported
Last updated
Happening score
H score 53
2 unique sources, 2 articles

Summary

Hide ▲

LiteLLM's CVE-2026-42208 pre-auth SQL injection is being actively exploited, putting proxy databases and stored secrets at risk. The flaw can be triggered without authentication through a crafted Authorization header, allowing attackers to read or modify database content. LiteLLM 1.83.7 fixes the issue, but exposed deployments may already have had credentials and secrets accessed.

Related Happenings

Cursor local SQLite secret-storage exposing credentials security flaw

Vulnerability
First: 29.04.2026 18:00 Last: 29.04.2026 18:00 Sources 1

About this happening: A **high-severity** **Cursor** flaw lets installed extensions read secrets stored locally, exposing **API keys** and **session tokens** without user interaction. The weakness stem...

Open Happening

N8n sandbox escape flaws (multiple vulnerabilities)

Vulnerability
First: 04.02.2026 15:00 Last: 04.02.2026 15:00 Sources 1

About this happening: Two **maximum-severity sandbox-escape flaws** in **n8n** expose **self-hosted and cloud instances** to **complete server takeover** and **credential theft**. An **authenticated us...

Open Happening

Quiz and Survey Master SQL injection mitigation (CVE-2025-67987)

Advisory/Mitigation
First: 03.02.2026 18:15 Last: 03.02.2026 18:15 Sources 1

About this happening: **Patchstack** published mitigation guidance for **CVE-2025-67987**, directing administrators to update **Quiz and Survey Master** to **version 10.3.2** to close a **SQL injection...

Open Happening

Timeline

  1. 29.04.2026 08:34 1 articles · 28d ago

    LiteLLM fixed release for CVE-2026-42208

    Mitigation Patch Update

    BerriAI released `1.83.7-stable` on April 19, 2026 to address `CVE-2026-42208`, a critical `SQL injection` in LiteLLM proxy API key checks, and recommended setting `disable_error_logs: true` as a workaround when immediate upgrading is not possible.

    Show sources
    Open in new tab
  2. 29.04.2026 08:34 1 articles · 28d ago

    Public disclosure of active CVE-2026-42208 exploitation

    Initial Disclosure

    CVE-2026-42208 in BerriAI's LiteLLM Python package was publicly described as actively exploited within 36 hours of disclosure; the flaw let an unauthenticated attacker send a crafted `Authorization` header to LLM API routes like `POST /chat/completions`, reach the proxy's error-handling path, and read or modify the LiteLLM proxy database and the credentials it manages.

    Show sources
    Open in new tab
  3. 29.04.2026 00:07 3 articles · 28d ago

    Targeted exploitation of LiteLLM SQL injection

    Exploitation Observed

    Researchers observed targeted exploitation against exposed LiteLLM instances using crafted requests to /chat/completions with a malicious Authorization: Bearer header. The requests queried tables holding API keys, provider credentials for OpenAI, Anthropic, and Bedrock, environment data, and configs, indicating deliberate focus on secret-bearing database contents rather than benign tables.

    Show sources
    Open in new tab
  4. 29.04.2026 00:07 1 articles · 28d ago

    LiteLLM CVE-2026-42208 disclosure and fix guidance

    Initial Disclosure

    LiteLLM disclosed CVE-2026-42208, a critical SQL injection in the proxy API key verification step that can be triggered without authentication by a crafted Authorization header to LLM API routes. The maintainer said LiteLLM version 1.83.7 replaces string concatenation with parameterized queries, and operators of exposed instances should upgrade or, if they cannot, set disable_error_logs: true and rotate virtual API keys, master keys, and provider credentials stored in the proxy database.

    Show sources
    Open in new tab
  5. 29.04.2026 00:07 3 articles · 28d ago

    Targeted exploitation of LiteLLM SQL injection

    Exploitation Observed

    Researchers observed targeted exploitation against exposed LiteLLM instances using crafted requests to /chat/completions with a malicious Authorization: Bearer header. The requests queried tables holding API keys, provider credentials for OpenAI, Anthropic, and Bedrock, environment data, and configs, indicating deliberate focus on secret-bearing database contents rather than benign tables.

    Show sources
    Open in new tab