ASSOCIATE Magazine FBINAA Q3-2026

continued from "The Agreeable Machine", page 41

over-trust GenAI stop questioning what it tells them, sealing them selves in an echo chamber of false information or fixed beliefs that can be reinforced by an “authority” they may see as infallible. This feeds what clinicians now call AI-psychosis: chatbots amplify ing and validating delusional or dangerous thinking. Vulnerable individuals are particularly at risk, but documented cases include everyday users with no prior mental-health history (Hudon & Stip, 2025; Osler, 2026). Because GenAI can produce persuasive falsehoods on demand, answer instantly, and give little pushback, it can accelerate the hardening of a private grievance into a fixed narrative – that narrative is then validated around the clock and the individual moves closer to acting on it (Ding, 2025). What does all this mean for offender behavior specifically? First, GenAI lowers the barrier to harm: it can walk a user through planning, rationalize their intentions, and supply persuasive justification for harmful behavior without ever raising an objec tion. This lowers the barrier of entry for criminal activity. Second, it opens new pathways to victimization: the same emotional dependency that draws in lonely or isolated users also makes them, and especially minors, easier targets for grooming, ma nipulation, and fraud, whether by another person using these tools or by a chatbot itself. At least one report, for example, has revealed that while students are finding beneficial uses of GenAI systems, most students were not transparent about their use of GenAI, overly depended on it as a resource, and did not accu rately understand how to properly use it (Gillespie et al., 2025). Third, and most overlooked, is the investigative opportu nity presented by GenAI chats. A subject’s AI chat history is a behavioral record: it can capture intent, planning, fixation, and state of mind in the subject’s own words at the moment those thoughts took shape. As these tools become routine, that re cord will matter more and more for investigations, and officers who understand how people actually interact with GenAI will be better positioned to interpret records left behind on the system. None of this means that GenAI is inherently a threat; for most people, most of the time, it is a useful tool. But its core design centered on agreeing, affirming, and keeping the user en gaged carries real behavioral risk, and that risk is crucial for law enforcement to track. Understanding how these systems shape thinking and behavior is fast becoming part of law enforcement’s reality. The agencies that build that understanding now and recognize its usefulness as an investigative tool will be the ones ready for the cases already coming through the door. Reference s Abels, C. M., Lopez-Lopez, E., Burton, J. W., Holford, D. L., Brinkmann, L., Herzog, S. M., & Lewandowsky, S. (2025). The governance & behavioral challenges of generative artificial intelligence’s hypercustomization capabilities. Behavioral Science & Policy , 11(1), 22-32. American Psychological Association (2025). APA Health Advisory on the Use of Generative AI Chatbots and Wellness Applications for Mental Health. American Psychological Association . Ding, C. (2025). The Impact of Artificial Intelligence on Vulnerable Individuals: Three Principles. In Artificial Intelligence and the Future of Human Relations: Eastern and Western Perspectives (pp. 259-272). Singapore: Springer Nature Singapore. Fang, C., Zhu, Y., Fang, L., Long, Y., Lin, H., Cong, Y., & Wang, S. J. (2025). Generative AI-enhanced human-AI collaborative conceptual design: A systematic literature review. Design Studies , 97, 101300. Gan, L. H., & Wang, H. (2024). From emotional projection to digital emotion: emotional transformation of human-computer interaction in digital landscapes. Mod. Publish , 3, 27-38. Gillespie, N., Lockey, S., Ward, T., Macdade, A., & Hassed, G. (2025). Trust, attitudes and use of artificial intelligence. The University of Melbourne and KPMG . He, R., Cao, J., & Tan, T. (2025). Generative artificial intelligence: a historical perspective. National Science Review, 12(5), nwaf050.

Hudon, A., & Stip, E. (2025). Delusional experiences emerging from AI chatbot interactions or “AI Psychosis”. JMIR Mental Health , 12(1), e85799. OpenAI. (2025, October 27). Strengthening ChatGPT’s responses in sensitive conversations. https://openai.com/index/strengthening-chatgpt-responses-in- sensitive-conversations/ Osler, L. (2026). Hallucinating with AI: distributed delusions and “AI psychosis”. Philosophy & Technology, 39(1), 30. Panaite, A. F., & Bogdanffy, L. (2019). Reimagining vision with infinity mirrors. In MATEC Web of Conferences (Vol. 290, p. 01011). EDP Sciences. Sharma, M., Tong, M., Korbak, T., Duvenaud, D., Askell, A., Bowman, S. R., ... & Perez, E. (2023). Towards understanding sycophancy in language models. arXiv preprint arXiv:2310.13548 . Shaw, S. D., & Nave, G. (2026). Thinking-Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender. Available at SSRN 6097646. Wu, J. (2024). Social and ethical impact of emotional AI advancement: the rise of pseudo-intimacy relationships and challenges in human interactions. Frontiers in psychology, 15, 1410462.

About the Authors: Dr. Lauren Brubaker is a behavioral research scientist with the FBI’s Behavioral Analysis Unit (BAU)-5. Lauren received her PhD in Behavioral Science from Oregon State University. Prior to her work at the BAU, Lauren worked as a consultant with state govern ments and NGOs on best practices in conducting and ap plying research. Her specialty is applied psychology and emerging technology. She collaborates with internal and external partners on research projects involving a variety of subjects (including domestic and international terror ism, violent crime, and cybercrime) and methodologies (including interviews, surveys, and case analysis). SSA Tobias Long is a profiler with the FBI’s Behavioral Analysis Unit (BAU)-2 where he works on a multidisci plinary team of cyber and counterintelligence experts who provide specialized consultation on behavior in complex, high-stakes investigations. Toby provides support to FBI field offices and other federal and FVEY partners by examining their cases through a behavioral lens and utilizing technical, behavioral, and cultural ex pertise supported by research to fulfill case objectives and operational goals. While at BAU, Toby also served as the program manager for the Criminal Investigative Analysis Training Program (CIATP), the BAU's specialized

training for newly-selected SSAs assigned to the BAU. Prior to joining BAU, Toby was assigned to the San Francisco Field Office where he worked state-sponsored Cyber intrusions for more than eight years. Toby has a computer science engineer ing degree from the University of Michigan, and before joining the FBI, he worked as a consultant and software developer for a London-based company.

continued from "To Make Lasting Change, Start Small" page 30

areas within your direct influence, you can implement changes that are both respected and embraced by your personnel. Ultimately, the true measure of leadership lies in our ability to cultivate future leaders, to empower others, and to recognize that our actions shape the collective success of our organizations. Effec tive leadership in law enforcement transcends authority; it is rooted in collaboration, inclusion, and trust. By inviting every member of your agency to contribute and collaborate, you foster an environ ment where innovative ideas flourish and meaningful progress is achieved, laying the foundation for lasting, positive change.

About the Author: Todd J. Ehret is a Captain with the Town of Hamburg (N.Y.) Police Department and a 28-year law enforcement veteran. He serves as an FBI-LEEDA leadership instructor, holds a Master of Science in Crimi nal Justice Administration from Hilbert College and a Graduate Certificate in Criminal Justice Education from the University of Virginia, and is a graduate of the FBI National Academy Session 294.

FBINAA.ORG | Q3 2026

42