Israeli startup 42AI emerged from stealth with $62.5 million in seed funding to develop what it calls a "behavioral world model" designed to predict how individuals will act in new situations. The company is currently deployed at roughly 10 large brokerages, where it analyzes customer behavior across transactions, support calls, onboarding processes, and trading activity to forecast trader responses to volatile market conditions.
The model works by converting observations of customer interactions into mathematical representations of psychological traits, then feeding those into a transformer-based system that generates probability distributions for potential decisions under different scenarios. CEO Amit Gal-Or said the approach combines psychological research with neuroscience and genetics, drawing on datasets covering more than 2 million people from government and biobank sources, plus almost 30 million people from customer records.
42AI claims its model achieves 98% accuracy in predicting whether trading flows fall within predicted ranges, and an F1 score of 0.8 to 0.85 for predicting trading direction and volume. Gal-Or said one brokerage generated approximately $30 million in additional value through dynamic hedging using the system. The company generates more than "seven figures" in revenue and charges customers either a flat fee or a percentage of platform gains.
The $62.5 million seed round came from the founders and Infinity Group over three years. Infinity was founded by Amit Gal-Or's father, Amir Gal-Or, who serves as 42AI's chairman and co-founder. Gal-Or said the round wasn't entirely raised from external investors and the company plans additional fundraising.
CTO Ehud Ezra acknowledged concerns about potential misuse, calling it "a valid concern" when raised. The founders pointed to regulations, monitoring, and an internal ethics committee as safeguards, though they did not detail what would prevent customers from using the technology to gain advantages over profiled individuals.
Beyond finance, 42AI is exploring applications in gaming, fraud prevention, cybersecurity, and helping robots and AI agents understand people, though Ezra noted the model requires retraining on rules and context for each new domain.