---
id: "CVE-2024-8502"
kind: "vulnerability"
aliases: ["CVE-2024-8502", "GHSA-9w5h-67gf-xvv8"]
title: "AgentScope Deserialization Vulnerability"
severity: "critical"
cvss: 9.8
epss: 0.01844
exploited: false
status: "open"
category: "ai-tooling"
products: [{"name": "agentscope", "fixed": [], "affected": "<= 0.0.6a3", "ecosystem": "pypi"}]
ecosystems: ["pypi"]
tags: ["agentscope", "ai-agents", "ai-tooling", "cwe-502", "python", "self-hosted"]
not_affected_if: ["newer-version", "no-rpc-server"]
published_at: "2025-03-20T10:15:42+00:00"
first_seen_at: "2026-10-26T20:56:17+00:00"
updated_at: "2026-10-26T21:20:32+00:00"
sources: [{"source": "ghsa", "url": "https://github.com/advisories/GHSA-9w5h-67gf-xvv8", "first_seen_at": "2026-10-26T20:56:17+00:00"}, {"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2024-8502", "first_seen_at": "2026-10-26T21:20:32+00:00"}]
---

# AgentScope Deserialization Vulnerability

AgentScope's RPC agent server calls dill.loads on serialized input in create_agent, so untrusted data is deserialized as Python objects. An attacker who can reach that endpoint can execute arbitrary commands on the server.

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

## Affected

- pypi:agentscope affected <= 0.0.6a3 fixed ?

## How to check

- Run `pip show agentscope` and check whether the version is `0.0.6a3` or older.
- Check whether the RpcAgentServerLauncher / agent server is running in your deployment (unknown whether it is exposed by default).

## Fix

- Upgrade the `agentscope` package to a version newer than `0.0.6a3` (the fixed version is unknown from the sources; check GHSA-9w5h-67gf-xvv8).
- Until upgraded, do not expose the agent server to untrusted networks.

## Timeline

- 2026-10-26T20:56:17+00:00 ghsa https://github.com/advisories/GHSA-9w5h-67gf-xvv8
- 2026-10-26T21:20:32+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2024-8502

## From the sources

A vulnerability in the RpcAgentServerLauncher class of modelscope/agentscope v0.0.6a3 allows for remote code execution (RCE) via deserialization of untrusted data using the dill library. The issue occurs in the AgentServerServicer.create_agent method, where serialized input is deserialized using dill.loads, enabling an attacker to execute arbitrary commands on the server.

---

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