CVE-2026-5970Public exploit

Arbitrary code execution via unsafe benchmark exec

Published Apr 9, 2026 · Updated Apr 9, 2026

Code injection in MetaGPT 0.8.0 and 0.8.1 allows attackers to execute arbitrary Python code through poisoned benchmark inputs. HumanEvalBenchmark and MBPPBenchmark check_solution pass LLM-generated solutions and test cases to Python exec after ineffective sanitization. An attacker must influence an LLM response or evaluation dataset; the aflow optimization loop then runs the payload on the MetaGPT host without review.

CVSS severity9.8
Critical
EPSS probability0.39%
Next 30 days · Sep 16, 2026
Known exploitationUnconfirmed
Based on sourced intelligence
Hinoki checkNot available
Coverage for this vulnerability

See if you're affected

Explore vulnerability checks for your environment with Hinoki.

Book a demo

Summary

What happened

Code injection in MetaGPT 0.8.0 and 0.8.1 allows attackers to execute arbitrary Python code through poisoned benchmark inputs. HumanEvalBenchmark and MBPPBenchmark check_solution pass LLM-generated solutions and test cases to Python exec after ineffective sanitization. An attacker must influence an LLM response or evaluation dataset; the aflow optimization loop then runs the payload on the MetaGPT host without review.

The record

CVE
CVE-2026-5970
Published
Apr 9, 2026
Updated
Apr 9, 2026
Vendor
Foundation Agents
Product
MetaGPT
Classifications
CWE-94, CWE-74, T1059.006
Attack vector
local
Privileges
unauthenticated

Timeline

How it unfolded

  1. Apr 9, 2026CVE publishedPublication date reported by the CVE source.
  2. Apr 9, 2026Record updatedLatest update available in the CVE record.

Exploitability

Present is not the same as exploitable

Compare your product and version with the public record. A matching version still requires validation against your environment.

Is a vulnerable build present?

Compare these published version ranges with your installed build and any vendor patches.

  1. Affected versionversion=0.8.0
  2. Affected versionversion=0.8.1

What conditions does exploitation require?

Attack vectorlocal
Required privilegesunauthenticated

What is affected?

Foundation Agents · MetaGPTversion=0.8.0; version=0.8.1

Published CVSS scores

7.5VulDB CNAAV:N/AC:L/Au:N/C:P/I:P/A:P
9.8NIST NVDCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
7.3VulDB CNACVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L
5.5VulDB CNACVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P

CVSS describes severity. EPSS estimates exploitation probability.

Attacks

What attackers are doing with it

Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.

Daily unique IPsNo honeypot observations are available for this CVE in the selected window.

No observations available

Sep 9, 2026Sep 15, 2026
Latest reporting daySep 15, 2026
Latest daily unique IPsUnavailable
Prior 30-day averageUnavailable
SourceShadowserver honeypots (KEV)
Vectorlocal
Privilegesunauthenticated
Known exploitationUnconfirmed
Public exploitPublished

Weakness, pattern, technique

CWE-94Improper Control of Generation of Code ('Code Injection')
CWE-74Improper Neutralization of Special Elements in Output Used by a Downstream Component ('Injection')
T1059.006Command and Scripting Interpreter: Python

Public exploit references

Labels summarize the accepted research assessment. They do not indicate a test against your environment.

Technologies

Your stack

See the directory against your own environment.

Your stack

Check the software in your environment

Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.

Book a demo