---
id: "CVE-2026-115802"
kind: "vulnerability"
aliases: ["CVE-2026-115802"]
title: "InternAgent: LoRA loading lets remote attackers pull untrusted model code"
severity: "medium"
cvss: 5.0
epss: 0.00486
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-tooling", "internagent", "python"]
not_affected_if: ["unknown"]
published_at: "2026-10-24T07:16:46+00:00"
first_seen_at: "2026-10-24T08:17:08+00:00"
updated_at: "2026-10-25T12:38:49+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-115802", "first_seen_at": "2026-10-24T08:17:08+00:00"}, {"source": "epss", "url": "https://api.first.org/data/v1/epss?cve=CVE-2026-115802", "first_seen_at": "2026-10-25T12:38:49+00:00"}]
---

# InternAgent: LoRA loading lets remote attackers pull untrusted model code

InternAgent 1.5.0 has a weakness in the train function of tasks/AutoChem/code/experiment.py, where the base_model_name_or_path argument is used to load functionality from an untrusted source. A remote attacker who supplies a crafted value can get their own code included and run.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- Check whether the installed InternAgent version is `1.5.0` in the project's package or release metadata.
- Search the code for `base_model_name_or_path` in `tasks/AutoChem/code/experiment.py` and see whether it reaches a model loading call with remote code trust enabled.

## Fix

- unknown

## Timeline

- 2026-10-24T08:17:08+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-115802
- 2026-10-25T12:38:49+00:00 epss https://api.first.org/data/v1/epss?cve=CVE-2026-115802

## From the sources

A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument base_model_name_or_path causes inclusion of functionality from untrusted control sphere. The attack may be initiated remotely. The exploit has been made available to the public and could be used for attacks.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_