42AI exits stealth with $62.5M and a world model built to predict human behavior

Israeli startup 42AI emerged from stealth on Monday with $62.5 million in seed funding and an unusual pitch: a “behavioral world model” that it says can predict what a person will do next, even in a situation it has never seen. The company is initially running the system at a handful of financial brokerages, where it says the model can anticipate how individual traders will respond to volatile market conditions.

CEO and co-founder Amit Gal-Or described the premise to TechCrunch as fundamentally probabilistic: “Behavior is not deterministic; it's constantly probabilistic.” He argues that people carry underlying psychological tendencies that shape their decisions even when their circumstances change, and that 42AI's job is to measure those tendencies and turn them into data a model can read.

Rather than extrapolating from a person's past actions, 42AI says it tries to infer internal motivations. It watches how people react to seemingly unrelated events — for a trader, completing KYC checks, onboarding, support calls, money transfers or placing trades. “When the person is talking to support or sales in the initial process … we're looking at the sentiment, we're looking at the voice, we're looking at a lot of interactions,” CTO and co-founder Ehud Ezra said.

Those observations are converted into mathematical representations of traits and fed into a transformer-based world model that lays out probabilities for possible decisions under different circumstances; a separate language model generates explanations. Training rewards are structured so that simulations preserve a person's underlying tendencies, Ezra said — if someone is usually risk-averse and the model suddenly treats them as a risk-taker, “the target is off.”

42AI says it spent three years on the approach, combining psychology, neuroscience and genetics with advice from psychology and PTSD experts and 2011 Nobel laureate in chemistry Dan Shechtman. It began with psych assessments and behavioral scenarios for thousands of people, then added government and biobank datasets covering more than 2 million; Gal-Or says customer data now spans almost 30 million people. The company claims 98% accuracy in predicting market trades, though Gal-Or clarified that measures whether trading flows fall inside a predicted range rather than whether each trade was called correctly, and says F1 for trading direction and volume sits at 0.8 to 0.85.

Commercially, 42AI says it works with 10 large brokerages that use it for liquidity management, margin requirements and hedging exposure, paying either a flat fee or a share of the value created. Gal-Or says revenue is above “seven figures” and claims one brokerage generated roughly $30 million in additional value through dynamic hedging. The $62.5 million came from the founders and Infinity Group over three years rather than entirely from outside investors; Infinity was founded by his father, Amir Gal-Or, who is 42AI's chairman and a co-founder. The company now plans to raise more.

The founders are candid about limits. Gal-Or said there is “a natural ceiling because you really don't have access to everything that a person experiences,” and Ezra said accuracy suffers when the model misses situational context, such as a trader acting on a colleague's advice. Expanding into gaming, fraud prevention, cybersecurity — and eventually helping robots and AI agents understand people — would require retraining rules and context for each domain.

The more uncomfortable question is what happens when a brokerage can anticipate its customers' behavior. Ezra called it “a valid concern” when TechCrunch raised the risk, and the founders pointed to regulation, monitoring and an internal ethics committee, without explaining what would stop customers from using the technology to gain an advantage over the very people it profiles.