42AI, a Tel Aviv startup building behavioral world models for finance, came out of stealth Monday with $62.5 million in seed funding. It said 10 large brokerages already use its model to predict client trades up to 24 hours in advance.
The money did not arrive the way most venture rounds do. TechCrunch reported on Nov. 2 that the capital built up over three years, mostly from the founders and from Infinity Group, the investment firm founded by co-founder and chairman Amir Gal-Or.
Brokerages as the First Customers
Financial brokerages are 42AI's starting point. The company told TechCrunch that clients use the model to manage liquidity, set margin requirements and hedge exposure. It also said annual recurring revenue, the yardstick for subscription-style businesses, is already in seven figures, though it declined to give an exact number. Pricing is either a flat fee or a share of the value the model generates.
On that last point, the company said one brokerage captured roughly $30 million in added value through dynamic hedging. The customer was not named. The figure comes from 42AI itself and has not been independently audited. The same goes for the 10 brokerages, none of which have been identified.
The company's website says the model forecasts each client's trades by instrument, along with direction and volume, and that API response times run from 1 to 12 milliseconds depending on the endpoint.
Inside the Model
According to TechCrunch, the system draws on onboarding and know-your-customer data, recordings of support and sales calls (voice, sentiment, how often a client phones in), money transfers and trading activity, all lined up against market conditions. That material is converted into mathematical "trait" representations and passed to a transformer-based behavioral world model. A separate language model writes the explanations.
42AI cites two accuracy numbers, and they measure different things. The 98% figure describes whether trading flows land inside a predicted range, not whether any single trade is called correctly, TechCrunch reported. For direction and volume of individual trades in live brokerage deployments, the company reports an F1 score (a blend of precision and recall, where 1.0 is perfect) of 0.80 to 0.85. As of Nov. 3, there was no third-party audit of either claim.
"Behavior is not deterministic; it's constantly probabilistic," CEO Amit Gal-Or told TechCrunch.
CTO Ehud Ezra framed it in more personal terms. "A week from now, you might be a different person, but your tendencies will always play out … to a probability or distribution," he said.
The People Behind It
Ezra's About page describes him as an AI leader with 17 years in the field and a stint as CTO of two unicorns. Amit Gal-Or is a serial entrepreneur and investor who says his ventures have produced tens of millions in profit. The advisers carry some weight: Dan Shechtman, who won the 2011 Nobel Prize in Chemistry, is a senior advisor, and Ariel Goldstein of the University of Cambridge advises on neuroscience research. Roi Reichart leads cognitive AI, and Eyal Fruchter leads behavioral research.
Ezra also co-wrote a Feb. 19, 2026 arXiv paper, "Decoding the Human Factor," which introduces a Large Behavioral Model fine-tuned to predict individual strategic choices from trait profiles. The paper reports gains over base Llama-3.1-8B and compares results with Claude 4.5 Sonnet. The evaluation was controlled and scenario-based, and the authors acknowledge limits, including a convenience sample and hypothetical scenarios. It says little, in other words, about how the system performs in a live brokerage.
A Crowded Field, a Narrower Claim
Investors have been generous to companies that model human behavior. Simile raised a $200 million Series B at a $2 billion valuation on July 30, five months after emerging from stealth with a $100 million Series A. humans& raised a $480 million seed in January, and Aaru reached a $1 billion headline valuation on its Series A in December 2025. TechCrunch also wrote on Oct. 6 about Mirror Particle, which is building a world model of human behavior for market and brand research.
Most of those companies, TechCrunch noted, work with synthetic users or research cohorts. Its reporting sets 42AI apart by its claim to predict at the level of the individual, inside live brokerage operations.
The "world model" label itself invites some eye-rolling. Alexandre LeBrun, CEO of Yann LeCun's AMI, which raised about $1.03 billion in March, told TechCrunch: "In six months, every company will call itself a world model to raise funding."
Regulatory Questions
A tool that anticipates what customers will do next raises obvious concerns. TechCrunch reported that brokerages could end up with an informational edge over their own clients. The founders pointed to regulation, monitoring and an internal ethics committee, but did not detail specific safeguards.
"The model never knows who the person is," Amit Gal-Or said. Pressed on misuse, he answered: "We are very proud that actually we are part of the solution; we are the solution to this problem."
Regulators are paying close attention to AI in finance. The Bank of England's Oct. 30 report found that 86% of respondents use AI, up from 75% in 2024, with a median of 43 AI systems in production at international banks and broker-dealers. The Financial Conduct Authority's October publication on good and poor practice stresses fairness, transparency and explainability, along with the Consumer Duty. Under the EU AI Act, transparency rules took effect Aug. 2, 2026, and high-risk obligations phase in by Dec. 2, 2027 for stand-alone systems and Aug. 2, 2028 for embedded ones. Whether a behavior-prediction tool for brokerages counts as high-risk depends on how it is used, and firms have to make that assessment themselves.
What Comes Next
Finance comes first, 42AI says. The company is exploring gaming, fraud prevention and cybersecurity, and its stated longer-term goal is to help robots and AI agents understand people, though each sector would require domain adaptation, per TechCrunch.
The funding structure carries its own implication: with the money coming largely from the founders and Infinity Group rather than a conventional outside syndicate, there is no announced external investor group that has vetted the company's claims. Independent validation of the 98% and F1 figures, and brokerage customers willing to be named, would be the clearest test of whether the model holds up.
Editor's note: Amir Gal-Or, the chairman, and Amit Gal-Or, the CEO, are two different people and co-founders of 42AI.
