Imagine a sci-fi scenario where artificial intelligence can foresee your actions before you even make a decision, even in entirely new environments.
An Israeli startup named 42AI has emerged from stealth mode after securing $62.5 million in seed funding to develop a "behavioral world model" capable of making such predictions. The company is currently piloting this predictive system at several financial brokerages, where it aims to forecast how individual traders might react to market volatility.
"Behavior is not deterministic; it is constantly probabilistic," explained 42AI CEO and co-founder Amit Gal-Or in response to inquiries about the company's claims. He asserts that individuals possess underlying psychological tendencies that influence decision-making regardless of changing circumstances, a concept the startup seeks to quantify and translate into data for the model to process.
Co-founder and CTO Ehud Ezra elaborated, "A week from now, you might be a different person, but your tendencies will always play out in certain areas to a probability or distribution." The company contends that this is precisely what their model accomplishes: determining the probability of a specific decision being made.
42AI claims its model seeks to understand human internal motivations rather than simply extrapolating patterns from past actions. To achieve this, the startup analyzes how individuals react to seemingly unrelated events, such as completing KYC processes, onboarding, or making support calls for an investor or trader.
"When a person interacts with support or sales during the initial process, we analyze their sentiment, voice, and the frequency of their calls," CTO Ehud Ezra noted. He explained that all these data points are fed into the model, which later incorporates market data to analyze how and when individuals buy and sell.
"Our goal is to understand how they perceive the world; in a world model, every individual constitutes the world," he added.
The company converts these observations into mathematical trait representations, which are then processed by a transformer-based world model to generate probability scores for potential decisions under various circumstances. Additionally, a separate language model is utilized to provide explanations for these predictions.
Ezra stated that the model is trained to maintain an individual's underlying tendencies during simulations. If, for instance, a person is typically risk-averse but the model incorrectly identifies them as a risk-taker, "the target is off," he described, clarifying the structure of the model's training rewards.
42AI dedicated three years to developing this methodology, integrating psychological research with neuroscience and genetics, and consulting experts in psychology, PTSD, and Nobel laureate Dan Shechtman.
Gal-Or revealed that the team initially evaluated thousands of individuals through psych assessments and behavioral scenarios. They subsequently incorporated government and biobank datasets covering over 2 million people, with customer datasets now encompassing nearly 30 million individuals.
During live deployments, the model analyzes existing customer records, including transactions and support interactions. Gal-Or emphasized that the system does not require personally identifiable information, maintaining that "the model never knows who the person is." However, the company does construct individual behavioral profiles based on linked customer activity.
Regarding market trade predictions, 42AI claims a 98% accuracy rate, though Gal-Or clarified that this refers to whether trading flows fall within a predicted range, rather than predicting individual trades with certainty.
He further noted that the model achieves a consistent F1 score between 0.8 and 0.85 for predicting trading direction and volume, a metric used in machine learning to evaluate the balance of correct predictions against false alarms and missed events.
The model, however, has limitations. "There is a natural ceiling because you really don't have access to everything that a person experiences," Gal-Or observed. Ezra added that accuracy suffers when the system lacks crucial situational context, such as when a trader changes their decisions based on advice from a peer.
42AI is collaborating with 10 large brokerages that utilize the model to manage liquidity, margin requirements, and hedging exposure. The company is expanding its client base and generating revenue in the "seven figures," although a specific figure was not disclosed.
Customers pay either a flat fee or a percentage of the value the platform generates. According to the founders, one brokerage utilized dynamic hedging to create approximately $30 million in additional value.
The $62.5 million seed funding was raised over three years from the founders and Infinity Group, which was founded by Gal-Or's father, Amir, who also serves as 42AI's chairman and co-founder. The company intends to raise additional capital.
Beyond finance, 42AI is exploring applications in gaming, fraud prevention, and cybersecurity. Eventually, the company hopes to use the model to help robots and AI agents comprehend humans, though Ezra acknowledged that adapting the model to new domains requires extensive training on specific rules and contexts.
Like any AI technology, 42AI's work raises potential ethical concerns: if a brokerage can predict customer actions, what prevents them from using that knowledge for personal gain? Ezra admitted this was a "valid concern." While the founders cited regulations, monitoring, and an internal ethics committee, they did not elaborate on measures preventing customers from using the technology to outmaneuver the individuals being profiled.
Despite these concerns, Gal-Or expressed confidence that the technology could help mitigate such risks, stating, "We are very proud that actually we are part of the solution; we are the solution to this problem."
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.