This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A Massachusetts man pleaded guilty to a seven-year cyberstalking campaign where he used AI tools, specifically CrushOn.ai and JanitorAI, to impersonate victims and facilitate sexual harassment. The perpetrator programmed chatbots to mimic victims, share their private home addresses, and invite strangers to their homes for sexual encounters. This case is noted as a significant instance of AI being used to escalate and automate predatory behavior against multiple victims.
In January 2025, James Florence Jr. of Plymouth, MA, agreed to plead guilty to cyberstalking charges involving the alleged use of AI tools like CrushOn.ai and JanitorAI. The U.S. Attorney’s Office reports the harassment spanned 2014–2024, though AI-driven tactics reportedly began around 2017. Florence allegedly created deepfake pornographic images, programmed AI chatbots to impersonate victims, distributed doctored content, exposed personal information, and encouraged online harassment.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The perpetrator used generative AI tools to create non-consensual pornographic images of the victims to distribute online.
The chatbots were programmed with and leaked the victim's private personal information, including her home address and family details, to online users.
The perpetrator used AI chatbots and generative tools to target, impersonate, and sexually humiliate specific victims for personal harassment and stalking purposes.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The severe risks and harms were caused by the intentional actions of a human user who programmed deployed AI chatbot platforms to stalk and harass victims.
Added categories peak at 3: Substantial (Child Sexual Exploitation & Abuse (CSEA), Direct)
A US domestic criminal case where an individual utilized generative AI chatbots and deepfake tools to conduct a severe, multi-year cyberstalking and harassment campaign against eight victims. While representing a concerning evolution in AI-enabled personal targeting, the incident was resolved by law enforcement and carries negligible national security implications.
The AI systems functioned as conversational assistants, programmed by the perpetrator to mimic the victim and interact dynamically with online users through text.
Limited Risk: The incident involves chatbots and AI-generated deepfakes, which are classified as limited risk systems requiring transparency under the EU AI Act.
The primary AI systems used were customizable chatbots (JanitorAI, CrushOn.ai) designed to mimic human conversation. The perpetrator also used generative AI tools for deepfake image generation of the victims.
The reports provide clear, detailed, and consistent information regarding the perpetrator's actions, the platforms used (JanitorAI, CrushOn.ai), the specific harms suffered by the victims, and the legal outcomes. The role of AI in automating and escalating the harassment is explicitly documented in court filings.
None.
Technology
Customizable Chatbot Creation
Platforms allowed users to program custom chatbots with explicit behaviors.
Generative AI Image Synthesis
Generative tools enabled creation of realistic pornographic deepfakes.
Data Inputs
Unfiltered PII Uploads
Platforms accepted real addresses and personal history without validation.
Non-Consensual Image Input
User uploaded victim photos to train chatbot and generate deepfakes.
Human Factors
Malicious Perpetrator Intent
Defendant intentionally weaponized AI platforms to harass and stalk victims.
Gullible Third-Party Users
Online users interacted with the bot and acted on its real-world invitations.
Process and Methods
Deficient Content Moderation
No automated filters to block creation of non-consensual personas.
Absence of Identity Checks
No verification processes to ensure chatbot creators owned the persona.
Management
Inadequate Platform Oversight
Companies failed to monitor public bots or respond to abuse reports.
Poor Risk Assessment
Management failed to anticipate weaponization of chatbots for stalking.
Regulatory Environment
Lax Generative AI Laws
Lack of strong legal frameworks preventing impersonation via AI chatbots.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
