SSOAR Logo
    • Deutsch
    • English
  • English 
    • Deutsch
    • English
  • Login
SSOAR ▼
  • Home
  • About SSOAR
  • Guidelines
  • Publishing in SSOAR
  • Cooperating with SSOAR
    • Cooperation models
    • Delivery routes and formats
    • Projects
  • Cooperation partners
    • Information about cooperation partners
  • Information
    • Possibilities of taking the Green Road
    • Grant of Licences
    • Download additional information
  • Operational concept
Browse and search Add new document OAI-PMH interface
JavaScript is disabled for your browser. Some features of this site may not work without it.

Download PDF
Download full text

(external source)

Citation Suggestion

Please use the following Persistent Identifier (PID) to cite this document:
https://doi.org/10.14512/tatup.33.2.29

Exports for your reference manager

Bibtex export
Endnote export

Display Statistics
Share
  • Share via E-Mail E-Mail
  • Share via Facebook Facebook
  • Share via Bluesky Bluesky
  • Share via Reddit reddit
  • Share via Linkedin LinkedIn
  • Share via XING XING

Misuse of large language models: Exploiting weaknesses for target-specific outputs

Missbrauch von Large Language Models: Die Ausnutzung von Schwachstellen für zielgruppenspezifische Outputs
[journal article]

Klinkhammer, Dennis

Abstract

Prompt engineering in large language models (LLMs) in combination with external context can be misused for jailbreaks in order to generate malicious outputs. In the process, jailbreak prompts are apparently amplified in such a way that LLMs can generate malicious outputs on a large scale despite the... view more

Prompt engineering in large language models (LLMs) in combination with external context can be misused for jailbreaks in order to generate malicious outputs. In the process, jailbreak prompts are apparently amplified in such a way that LLMs can generate malicious outputs on a large scale despite their initial training. As social bots, these can contribute to the dissemination of misinformation, hate speech, and discriminatory content. Using GPT4-x-Vicuna-13b-4bit from NousResearch, we demonstrate in this article the effectiveness of jailbreak prompts and external contexts via Jupyter Notebook based on the Python programming language. In addition, we highlight the methodological foundations of prompt engineering and its potential to create malicious content in order to sensitize researchers, practitioners, and policymakers to the importance of responsible development and deployment of LLMs.... view less


Prompt Engineering in Large Language Models (LLMs) kann in Kombination mit externen Kontexten für Jailbreaks missbraucht werden, um bösartige Outputs zu erzeugen. Dabei werden 'jailbreak prompts' offenbar so verstärkt, dass LLMs trotz ihres ursprünglichen Trainings in großem Umfang bösartige Ausgabe... view more

Prompt Engineering in Large Language Models (LLMs) kann in Kombination mit externen Kontexten für Jailbreaks missbraucht werden, um bösartige Outputs zu erzeugen. Dabei werden 'jailbreak prompts' offenbar so verstärkt, dass LLMs trotz ihres ursprünglichen Trainings in großem Umfang bösartige Ausgaben generieren können. Als ‚social bots‘ können diese zur Verbreitung von Falschmeldungen, hate speech und diskriminierenden Inhalten beitragen. In diesem Artikel demonstrieren wir anhand von GPT4-x-Vicuna-13b-4bit von NousResearch die Effektivität von Jailbreak Prompts und externen Kontexten als Jupyter Notebook in der Programmiersprache Python. Darüber hinaus beleuchten wir die methodischen Grundlagen des Prompt Engineering und sein Potenzial, bösartige Inhalte zu generieren, um Forschung, Praxis und Politik für die Bedeutung einer verantwortungsvollen Entwicklung und Implementierung von LLMs zu sensibilisieren.... view less

Classification
Technology Assessment

Free Keywords
deep learning; jailbreak prompts; large language models; prompt engineering; transformers

Document language
English

Publication Year
2024

Page/Pages
p. 29-34

Journal
TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis / Journal for Technology Assessment in Theory and Practice, 33 (2024) 2

Issue topic
Malevolent creativity and civil security: The ambivalence of emergent technologies / Malevolente Kreativität und zivile Sicherheit: Die Ambivalenz neu entstehender Technologien

ISSN
2567-8833

Status
Published Version; peer reviewed

Licence
Creative Commons - Attribution 4.0


GESIS LogoDFG LogoOpen Access Logo
Home  |  Legal notices  |  Operational concept  |  Privacy policy
© 2007 - 2025 Social Science Open Access Repository (SSOAR).
Based on DSpace, Copyright (c) 2002-2022, DuraSpace. All rights reserved.
 

 


GESIS LogoDFG LogoOpen Access Logo
Home  |  Legal notices  |  Operational concept  |  Privacy policy
© 2007 - 2025 Social Science Open Access Repository (SSOAR).
Based on DSpace, Copyright (c) 2002-2022, DuraSpace. All rights reserved.