New AI company raises $62,5 million to predict human decisions

dev.to Machine Learning2026-11-03 00:410 views

Israeli startup 42AI has emerged from stealth with $62.5 million in seed funding, introducing a transformer-based behavioral world model designed to predict human decision probabilities and trader responses in financial markets using psychological and behavioral datasets.

According to its founders, Amit Gal-Or and Ehud Ezra, the system doesn't just look at what someone has done in the past. Instead, it tries to measure underlying psychological trends by analyzing various everyday actions, such as account creations, support calls, money transfers, and stock purchases.

The model took three years to develop, drawing on research in psychology, neuroscience and genetics, as well as input from experts and Nobel laureate Dan Shechtman. The system was trained on data from thousands of people, government records and biological data banks covering over 2 million people, while customer data now extends to nearly 30 million people.

Where is it already used?

Currently, 42AI is running its system at select brokerages, helping with cash management, collateral requirements and risk mitigation, which has already generated over $1 million in revenue. According to the company, live tests do not require any identification, as the system does not know who the user is.

↳

Limitations and ethical concerns

The founders themselves admit that the system's accuracy drops when important elements of everyday life are missing. For example, the system can't predict when a person will change their mind simply because they took advice from a friend or colleague.

At the same time, they acknowledge that there are legitimate concerns about the potential misuse of this type of knowledge. The company says it relies on regulations, audits and an internal ethics committee to address these issues.

Future plans include expanding the technology to electronic games, fraud prevention, system security, and helping robots and artificial intelligence programs better understand humans.