Skip to content
Search
paper2024ReviewedOpen access

GPT in Sheep's Clothing: The Risk of Customized GPTs

Sagiv Antebi, Noam Azulay, Edan Habler, Ben Ganon, Asaf Shabtai, Yuval Elovici

arXiv preprint

View resourcePDF

Abstract

Analyzes security risks of custom GPTs in the OpenAI GPT Store including prompt leakage, data exfiltration, and malicious GPTs.

Categories

Supply Chain AttacksPrompt InjectionModel Serving Security
#custom-GPTs#GPT-Store#supply-chain

Framework mappings

OWASP Top 10 for LLM Applications
  • LLM03Supply Chain
  • LLM07System Prompt Leakage
OWASP Top 10 for Agentic Applications
  • ASI04Agentic Supply Chain Compromise

Cite

@misc{antebi2024gpt,
  title = {{GPT in Sheep's Clothing: The Risk of Customized GPTs}},
  author = {Sagiv Antebi and Noam Azulay and Edan Habler and Ben Ganon and Asaf Shabtai and Yuval Elovici},
  year = {2024},
  eprint = {2401.09075},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2401.09075}
}