Malevolent Creativity: The Varied Use of Artificial Intelligence as an Emerging Threat
Download the full project summary.Project Summary
Artificial intelligence is moving beyond text-generating chatbots into embodied and autonomous forms such as humanoid robots, delivery systems, and driverless vehicles, and the homeland security enterprise has limited systematic evidence on how these systems could be turned to harmful ends or how they shape the thinking of the people who use them. Building on years of work on malevolent creativity, this project pursues four connected lines of inquiry.
Purpose/Objectives
This project gives DHS analysts and threat assessors an evidence-based account of how embodied and uncensored AI systems lower the barriers to generating and acting on novel attack ideas, along with the cognitive biases that lead practitioners to dismiss such ideas, supporting intelligence, emerging technology, and threat assessment work across components including I&A, S&T, and CBP.
Method
The project uses a mixed-methods design in four phases. Phase 1 conducts a systematic review of embodied, autonomous AI systems drawing on threat reports, case reports and court documents, and academic literature. Phase 2 runs a controlled 2x2 experiment varying task type, meaning malevolent or neutral, and AI type, meaning censored or uncensored, with analysis of participant prompts. Phase 3 runs a second experiment varying the AI system’s apparent personality across agreeableness, openness, and conscientiousness. Phase 4 produces a case study report on novel threat bias.
Outputs and Impact
- A systematic review and set of case briefs.
- A case report on novel threat bias documenting failures to assess emerging attack methods, with lessons learned for threat assessment practice.
Media Reporting
- These AI models are free, private, and will never say 'no' | NPR
- NPR article on uncensored AI models, featuring Sam Hunter
- House lawmakers get a chilling demo of ‘jailbroken’ AI | POLITICO
- Politico's coverage of the project's April 2026 workshop with D.C. lawmakers on uncensored and abliterated AI LLM models
- Recap Video | House Committee on Homeland Security
Reports and Publications
Examining Artificial Intelligence and Its Role in Malevolent Creativity and Innovation: How AI Can Enhance Novel Threats and AttacksThis report attempts to briefly summarize for the homeland security enterprise how terrorists could use generative AI.
Pixels and Propaganda: A Primer on Gaming, Gaming-Adjacent Platforms, and ExtremismThis report addresses the intersection of gaming and artificial intelligence with extremism, raising complex challenges that demand careful consideration.
The Metaverse as a Future Threat Landscape: An Interdisciplinary PerspectiveThis peer-reviewed report attempts to provide insight into the concept of "metaverse" and its potential use in extremist activity.
Research Team
Samuel Hunter, Ph.D.- Head of Strategic Operations at NCITE
- University of Nebraska at Omaha | Professor
- Department of Psychology
- Expertise: Malevolent creativity, leadership in organizations, creativity and innovation in organizations
- Head of IS&T Research Initiatives at NCITE
- University of Nebraska at Omaha | Assistant Professor
- School of Interdisciplinary Informatics
- Expertise: Collaboration technologies, decision support systems
Publications
- Malevolent Creativity as Parochial Altruism? Examining the Intergroup Bases of New and Harmful Ideas
- Creativity Training Needs Assessment for Homeland Security Enterprise: A Case for Creative Thinking
- Target Personification Influences the Positive Emotional Link Between Generating and Implementing Malevolently Creative Ideas