CVE-2026-26019

LangChain is a framework for building LLM-powered applications. Prior to 1.1.14, the RecursiveUrlLoader class in @langchain/community is a web crawler that recursively follows links from a starting URL. Its preventOutside option (enabled by default) is intended to restrict crawling to the same site as the base URL. The implementation used String.startsWith() to compare URLs, which does not perform semantic URL validation. An attacker who controls content on a crawled page could include links to domains that share a string prefix with the target, causing the crawler to follow links to attacker-controlled or internal infrastructure. Additionally, the crawler performed no validation against private or reserved IP addresses. A crawled page could include links targeting cloud metadata services, localhost, or RFC 1918 addresses, and the crawler would fetch them without restriction. This vulnerability is fixed in 1.1.14.

An adversary targeting a company that uses a LangChain-powered research agent for competitive intelligence identifies that the agent crawls industry news sites. The adversary compromises or creates a page on a domain sharing a prefix with the crawl target (e.g., target-news.com vs target-news.com.attacker.com) or embeds a link to http://169.254.169.254/latest/meta-data/iam/security-credentials/ on a page the crawler will visit. When the research agent's RecursiveUrlLoader processes the malicious page, it follows the link without validation, fetches the AWS metadata endpoint, and the response (containing temporary IAM credentials) is processed as content — potentially stored in the RAG database or logged. The attacker can then query the agent or access logs to extract the credentials and pivot to the cloud account.