# CVE-2026-119120: Crawl4AI untrusted model loading in CosineStrategy

Source: https://junglewise.ai/threats/cve-2026-119120-crawl4ai-from-0-9-0-before-0-9-5-fails-to-filter-the  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-29. Updated: 2026-10-29  
License: CC BY 4.0 with a link back

CVE-2026-119120 · Severity: high · CVSS 7.5 · Published 2026-10-29

Crawl4AI is a web scraping library used in Docker API deployments to extract and analyze web content. A flaw in versions 0.9.0 through 0.9.4 allows authenticated users to specify arbitrary machine learning models to download and load, which can lead to remote code execution on older PyTorch versions or exhaust server resources through disk and memory consumption.

Technologies: [Unclecode Crawl4AI](https://junglewise.ai/threats/technologies/unclecode-crawl4ai). Vendors: [Unclecode](https://junglewise.ai/threats/vendors/unclecode).

## Quick answers

**Is CVE-2026-119120 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-119120 being exploited in the wild.

**How severe is CVE-2026-119120?** High severity, with a CVSS score of 7.5, published 29 October 2026.

**Which products and versions are affected?** Unclecode Crawl4AI 0.9.0 through 0.9.4.

**Is there a fix for CVE-2026-119120?** Yes: fixed in 0.9.5. Until then: Reject crawler_config request bodies containing model_name at the reverse proxy. Upgrade PyTorch to 2.6 or newer. Pre-fill the HuggingFace cache with the model in use and set HF_HUB_OFFLINE=1. Use the default Crawl4AI image without cosine extraction extras if cosine similarity is not required. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-119120 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-119120?** Check whether you run Unclecode Crawl4AI at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix

- Fixed in 0.9.5

Reject crawler_config request bodies containing model_name at the reverse proxy. Upgrade PyTorch to 2.6 or newer. Pre-fill the HuggingFace cache with the model in use and set HF_HUB_OFFLINE=1. Use the default Crawl4AI image without cosine extraction extras if cosine similarity is not required.

## Technical details

The Docker API server fails to validate the model_name field in CosineStrategy extraction strategy requests, allowing authenticated clients to specify any HuggingFace model repository. When PyTorch is older than 2.6, a malicious pickle weight file in the repository executes arbitrary Python code via torch.load(). Even with current PyTorch, an attacker can force arbitrary model downloads, exhausting server disk and memory resources.

## Affected products

- unclecode Crawl4AI 0.9.0 through 0.9.4

## Timeline

- 2026-10-15: disclosed
- 2026-10-29: advisory
- 2026-10-29: patched: Version 0.9.5 released

## References

- [https://github.com/unclecode/crawl4ai/security/advisories/GHSA-f856-7xp9-w3cc](https://github.com/unclecode/crawl4ai/security/advisories/GHSA-f856-7xp9-w3cc)
- [https://www.vulncheck.com/advisories/crawl4ai-0.9.0-before-0.9.5-untrusted-model-loading-via-cosinestrategy-model-name](https://www.vulncheck.com/advisories/crawl4ai-0.9.0-before-0.9.5-untrusted-model-loading-via-cosinestrategy-model-name)

## Related threats

- [Crawl4AI blind SSRF in CacheValidator proxy bypass](https://junglewise.ai/threats/cve-2026-119122-crawl4ai-before-0-9-5-contains-a-blind-server-side-request)
- [Crawl4AI JavaScript injection in Docker API server](https://junglewise.ai/threats/cve-2026-119121-crawl4ai-from-0-9-0-before-0-9-5-contains-a-javascript-injection)
- [Crawl4AI denial of service in Docker API server](https://junglewise.ai/threats/cve-2026-119119-crawl4ai-before-0-9-5-contains-a-denial-of-service-vulnerability)
- [Crawl4AI improper certificate validation in Chromium launch](https://junglewise.ai/threats/cve-2026-119118-crawl4ai-before-0-9-5-contains-an-improper-certificate)
- [Crawl4AI inefficient regular expression complexity in Docker API](https://junglewise.ai/threats/cve-2026-115508-crawl4ai-0-7-0-before-0-9-5-contains-an-inefficient-regular)