CVE-2024-58340: LangChain security flaw enables

CISO Take

Any LangChain-based application running MRKL agents on version 0.3.1 or earlier is vulnerable to a DoS attack delivered via prompt injection — no authentication required. An attacker who can influence LLM output (e.g., through user-supplied prompts in a downstream app) can stall your agent service with a single crafted string. Patch to LangChain >0.3.1 immediately; if you cannot patch today, wrap MRKLOutputParser calls with a timeout and sanitize LLM output before parsing.

What is the risk?

CVSS 7.5 High with AV:N/AC:L/PR:N/UI:N is accurate for the worst case. Real-world exploitability requires the attacker to first achieve prompt injection against the target application — an increasingly realistic assumption for any public-facing LLM app. The exploit chain is two steps: (1) inject a crafted payload via user input, (2) the LLM reflects it and the app feeds it to the vulnerable parser. CPU exhaustion is the outcome, not code execution or data leakage, which limits blast radius. Organizations running MRKL/ReAct agent architectures at scale or in multi-tenant SaaS products face the highest exposure.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
LangChain	pip	—	No patch
139.8K OpenSSF 5.9 2.7K dependents Pushed 2d ago 24% patched ~156d to patch Full package profile →

Do you use LangChain? You're affected.

How severe is it?

CVSS 3.1

7.5 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 32% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Moderate

Exploitation Confidence

medium

○ CISA SSVC: Public PoC

○ Public PoC indexed (trickest/cve)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AV Network

AC Low

PR None

UI None

S Unchanged

C None

I None

A High

What should I do?

6 steps

PATCH

Upgrade langchain to the first version past 0.3.1 that includes the fixed regex; verify with pip show langchain.
WORKAROUND (if patch is not immediate): Wrap MRKLOutputParser.parse() calls with a signal-based or thread-based timeout (e.g., 2–5 seconds); raise a parsing error and abort on timeout.
INPUT HYGIENE

Truncate LLM output to a reasonable maximum length (e.g., 4 KB) before passing to the parser; reject outputs with suspicious repetitive patterns.
RATE LIMITING

Apply per-user/session rate limits on agent invocations to reduce DoS throughput.
DETECTION

Alert on sustained high CPU usage in agent worker processes; log parsing duration and alert on outliers >500 ms.
INVENTORY

Audit all internal and customer-facing apps that import langchain.agents.mrkl.

What does CISA's SSVC say?

Decision Track*

Exploitation poc

Automatable Yes

Technical Impact partial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

DoS Prompt Injection Framework Agent AML.T0010.001 - AI Software AML.T0029 - Denial of AI Service AML.T0034 - Cost Harvesting AML.T0043 - Craft Adversarial Data AML.T0049 - Exploit Public-Facing Application AML.T0051 - LLM Prompt Injection AML.T0051.000 - Direct AML.T0051.001 - Indirect

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 15 - Accuracy, robustness and cybersecurity Article 15 - Accuracy, Robustness, and Cybersecurity

ISO 42001

A.6.1.4 - AI Risk Assessment A.6.2.6 - AI system availability and resilience A.9.2 - AI Incident Handling A.9.3 - AI system risk treatment

NIST AI RMF

MANAGE-2.2 - Mechanisms to sustain the deployed AI system MAP-5.1 - Likelihood and Impact of AI Risks

OWASP LLM Top 10

LLM01:2025 - Prompt Injection LLM07:2025 - System Prompt Leakage / Insecure Output Handling LLM10:2025 - Unbounded Consumption

Frequently Asked Questions

What is CVE-2024-58340?

Any LangChain-based application running MRKL agents on version 0.3.1 or earlier is vulnerable to a DoS attack delivered via prompt injection — no authentication required. An attacker who can influence LLM output (e.g., through user-supplied prompts in a downstream app) can stall your agent service with a single crafted string. Patch to LangChain >0.3.1 immediately; if you cannot patch today, wrap MRKLOutputParser calls with a timeout and sanitize LLM output before parsing.

Is CVE-2024-58340 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2024-58340, increasing the risk of exploitation.

How to fix CVE-2024-58340?

1. PATCH: Upgrade langchain to the first version past 0.3.1 that includes the fixed regex; verify with `pip show langchain`. 2. WORKAROUND (if patch is not immediate): Wrap MRKLOutputParser.parse() calls with a signal-based or thread-based timeout (e.g., 2–5 seconds); raise a parsing error and abort on timeout. 3. INPUT HYGIENE: Truncate LLM output to a reasonable maximum length (e.g., 4 KB) before passing to the parser; reject outputs with suspicious repetitive patterns. 4. RATE LIMITING: Apply per-user/session rate limits on agent invocations to reduce DoS throughput. 5. DETECTION: Alert on sustained high CPU usage in agent worker processes; log parsing duration and alert on outliers >500 ms. 6. INVENTORY: Audit all internal and customer-facing apps that import langchain.agents.mrkl.

What systems are affected by CVE-2024-58340?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, LLM agentic pipelines, ReAct/MRKL agent workflows, multi-tenant LLM SaaS.

What is the CVSS score for CVE-2024-58340?

CVE-2024-58340 has a CVSS v3.1 base score of 7.5 (HIGH). The EPSS exploitation probability is 0.41%.

What is the AI security impact?

Affected AI Architectures

agent frameworksLLM agentic pipelinesReAct/MRKL agent workflowsmulti-tenant LLM SaaS

MITRE ATLAS Techniques

AML.T0010.001 AI Software

AML.T0029 Denial of AI Service

AML.T0034 Cost Harvesting

AML.T0043 Craft Adversarial Data

AML.T0049 Exploit Public-Facing Application

AML.T0051 LLM Prompt Injection

AML.T0051.000 Direct

AML.T0051.001 Indirect

Compliance Controls Affected

EU AI Act: Art. 15, Article 15

ISO 42001: A.6.1.4, A.6.2.6, A.9.2, A.9.3

NIST AI RMF: MANAGE-2.2, MAP-5.1

OWASP LLM Top 10: LLM01:2025, LLM07:2025, LLM10:2025

What are the technical details?

Original Advisory

LangChain versions up to and including 0.3.1 contain a regular expression denial-of-service (ReDoS) vulnerability in the MRKLOutputParser.parse() method (libs/langchain/langchain/agents/mrkl/output_parser.py). The parser applies a backtracking-prone regular expression when extracting tool actions from model output. An attacker who can supply or influence the parsed text (for example via prompt injection in downstream applications that pass LLM output directly into MRKLOutputParser.parse()) can trigger excessive CPU consumption by providing a crafted payload, causing significant parsing delays and a denial-of-service condition.

Exploitation Scenario

An attacker targets a public-facing AI assistant built on LangChain MRKL agents. They craft a user prompt designed to cause the underlying LLM to produce output containing a pathological string — for example, a long sequence of spaces or repeated characters that exploits the backtracking in the MRKL action-extraction regex (e.g., `Action: ` followed by thousands of repeated ambiguous characters). The application passes the LLM's raw output directly to MRKLOutputParser.parse() without sanitization. The regex engine enters catastrophic backtracking, pegging one CPU core at 100% for tens of seconds per request. An attacker automating dozens of such requests can exhaust worker threads and render the service unavailable for all users within minutes.

Weaknesses (CWE)

CWE-1333 Inefficient Regular Expression Complexity Primary

CWE-1333 — Inefficient Regular Expression Complexity: The product uses a regular expression with a worst-case computational complexity that is inefficient and possibly exponential.

[Architecture and Design] Use regular expressions that do not support backtracking, e.g. by removing nested quantifiers.
[System Configuration] Set backtracking limits in the configuration of the regular expression implementation, such as PHP's pcre.backtrack_limit. Also consider limits on execution time for the process.

Source: MITRE CWE corpus.