We Talked to an AI Trained on Jeffrey Epstein’s Emails. Here's What It Said
A software consultant, AIfredo Ortega, created an AI model called MechaEpstein-8000 by fine-tuning Qwen3-8B on Jeffrey Epstein’s released email correspondence. The model is available for free on Hugging Face, runs locally on consumer hardware, and has been downloaded over 33,000 times. MechaEpstein-8000 mimics Epstein’s email persona: terse, dismissive, using the term “goy” frequently, and ending messages with “Sent from my iPhone,” faithfully reproducing his communication style and typos. The AI answers questions as Epstein might have, denying wrongdoing, expressing elitism, emphasizing connections and power, and referring to high-profile names from Epstein’s network. It consistently deflects sensitive questions and judges people by usefulness. When pressed, the AI downplays controversial topics, pivots to small talk, and offers advice centered on relationships and assets over knowledge or crypto. The model is not designed for factual retrieval but for persona simulation based on Epstein’s self-image. The popularity of this AI, compared to similar models, underscores public curiosity about Epstein and his association with power amidst ongoing government disclosures.
