42AI has come out of stealth with $62.5 million in seed funding and a world model it says can predict human behaviour, TechCrunch reports. After three years of development, its first customers are brokerages. Its approach drew on advice from experts including the Nobel laureate Dan Shechtman, though how much he contributed is unclear.
The company says it works with 10 large brokerages and has customer datasets covering almost 30 million people, with revenue in the seven figures. It says one brokerage gained about $30 million in extra value from dynamic hedging. None of these figures is independently verified.
Its accuracy claims need careful reading.
It cites an F1 score of 0.8 to 0.85 for predicting trading direction and volume. A 98 percent figure was later clarified as measuring whether trading flows land within a predicted range, not whether each trade was called.
Chief executive Amit Gal-Or says behaviour is probabilistic and the model never knows who a person is. TechCrunch notes the company still builds individual behavioural profiles from linked customer activity.
Co-founder and chief technology officer Ehud Ezra called misuse a valid concern. The founders offered regulation, monitoring and an internal ethics committee, and did not explain what would stop a brokerage using the predictions against its own customers.
A large share of the seed came from the founders and a family-linked investment group rather than outside investors. Initial studies covered thousands of people, and the government and biobank datasets it used cover more than 2 million.
Every accuracy and revenue figure here is the company’s own, and the question of whose side the model works for has no answer yet.
