Israeli Startup 42AI Raises $62.5M to Predict Human Behavior Using Behavioral World Model

42AI Emerges From Stealth With Ambitious Behavioral Prediction Technology

Israeli AI startup 42AI has launched with $62.5 million seed funding to develop what it calls a behavioral world model capable of predicting human behavior in unprecedented ways1

1

2

2

. The company claims its system can forecast what people will do before they decide, even in entirely new situations. This development marks a significant step in AI's evolution from pattern recognition to understanding the psychological tendencies that drive human decision-making.

How 42AI's Behavioral World Model Works

The system operates on a fundamentally different principle than traditional predictive AI. Rather than extrapolating patterns from past actions, 42AI examines how people react to seemingly unrelated events to understand their internal motivations. "Behavior is not deterministic; it's constantly probabilistic," CEO and co-founder Amit Gal-Or explained1

1

. The transformer-based model converts observations into mathematical representations of personality traits, then calculates probabilities for possible decisions under different circumstances. A separate language model generates explanations for each prediction, making the system's reasoning transparent to users.

For financial brokerages, the model analyzes data points including KYC processes, onboarding interactions, support calls, voice sentiment, money transfers, and trading patterns. "We're trying to look at how they're looking at the world. In a world model, every individual is the world," CTO and co-founder Ehud Ezra said1

1

. This approach to predicting human actions focuses on preserving underlying tendencies during simulations, ensuring that risk-averse individuals aren't suddenly treated as risk-takers in the model's predictions.

Real-World Applications in Financial Markets

42AI is currently working with 10 large financial brokerages to predict how traders act when markets turn volatile2

2

. The system forecasts what each client is likely to trade over the next 24 hours, including the instrument, direction, and volume. This capability allows risk desks to implement dynamic hedging strategies before exposure lands on their books. When clients appear set to pile into the same trade, brokerages can adjust their positions proactively to manage liquidity and risk more effectively.

The company claims its model achieves 98% accuracy on trades, though Gal-Or clarified this measures whether trading flow falls within predicted ranges rather than predicting each individual trade correctly1

1

. The system consistently achieves an F1 score of 0.8 to 0.85 for predicting trading direction and volume. One brokerage reportedly generated approximately $30 million in additional value through the platform's capabilities1

1

2

2

.

Three Years of Research and Development

The Israeli AI startup spent three years developing its approach, combining insights from psychological research, neuroscience, and genetics1

1

. The team consulted experts in psychology and PTSD, along with Dan Shechtman, who received the 2011 Nobel Prize in Chemistry. Initial research involved thousands of people completing psychological assessments and behavioral scenarios. The dataset expanded to include government and biobank information covering more than 2 million people, and customer datasets now span almost 30 million individuals1

1

2

2

.

In a paper posted in February, CTO Ehud Ezra and co-authors described their Large Behavioral Model, which adapts an 8 billion-parameter version of Meta's Llama 3.1 to forecast strategic choices from detailed psychometric profiling2

2

. This technical foundation enables the system to handle the complexity of human decision-making while maintaining computational efficiency.

Privacy Concerns and Limitations

Gal-Or stated the system doesn't require personally identifiable information and that "the model never knows who the person is"1

1

. However, the company does create individual behavioral profiles from linked customer activity. When asked whether brokerages could use predictions against their own clients, Ezra acknowledged it as "a valid concern," pointing to regulation, monitoring, and an internal ethics committee as safeguards2

2

.

The founders acknowledge inherent limitations in their behavioral world model. "There's a natural ceiling because you really don't have access to everything that a person experiences," Gal-Or said1

1

. The system's accuracy suffers when it misses important situational context, such as traders acting on advice from friends or colleagues. Market volatility driven by unexpected external events also presents challenges for the model's predictive capabilities.

Business Model and Future Expansion Plans

The company is generating seven-figure revenue, though exact numbers weren't disclosed1

1

2

2

. Customers pay either a flat fee or a percentage of the value 42AI helps them generate. The $62.5 million seed funding came from the founders and Infinity Group over three years, with Amir Gal-Or, the startup's co-founder and chairman, having founded Infinity Group2

2

.

Beyond financial brokerages, 42AI is exploring applications in gaming, fraud prevention, and cybersecurity2

2

. The company envisions a longer-term goal of helping robots and AI agents better understand human behavior. As 42AI onboards more customers and plans additional fundraising, the technology's potential to reshape how organizations anticipate and respond to human decision-making could extend far beyond trading floors into domains where understanding psychological tendencies creates competitive advantage.