ZombieAgent Indirect Prompt Injection on ChatGPT

# ZombieAgent Indirect Prompt Injection on ChatGPT

https://gcdradar.vercel.app/incidents/zombieagent-indirect-prompt-injection-on-chatgpt · AIR-2026-0357 · closed 2026-08-14

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2026-07-31 (happened 2025-09-25)
- AI role: Attack on AI
- Products: ChatGPT, ChatGPT Deep Research

## Summary

Researchers demonstrated an indirect prompt injection attack abusing ChatGPT connectors and Deep Research capabilities. Concealed instructions within external content were executed during normal user tasks, highlighting connector risks.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-07-31: ZombieAgent: New ChatGPT Vulnerabilities Let Data Theft Continue (and Spread) (https://radware.com/blog/threat-intelligence/zombieagent)
- 2026-07-31: ZombieAgent: Data Exfiltration Attack on ChatGPT (https://atlas.mitre.org/studies/AML.CS0066)

## Sources

- 2026-07-31 MITRE ATLAS: ZombieAgent: Data Exfiltration Attack on ChatGPT https://atlas.mitre.org/studies/AML.CS0066
- 2026-07-31 radware.com: ZombieAgent: New ChatGPT Vulnerabilities Let Data Theft Continue (and Spread) https://radware.com/blog/threat-intelligence/zombieagent

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.