42AI Develops AI Model That Predicts Human Decisions

42AI has developed a behavioral model designed to predict what decision a person will make under new circumstances. The Israeli startup has emerged from stealth and announced that it has raised $62.5 million (approximately 106.3 million manats) in seed, or initial, funding. The system is currently being tested by several financial brokers. The goal is to calculate how each trader will behave when the market changes sharply.

42AI does not look only at past actions; it also seeks to measure the tendencies behind a person's decisions. The model examines data collected during customer identity verification (KYC), registration, support calls, money transfers, and stock purchases and sales. It also considers tone of voice, the number of calls, the timing of transactions, and market conditions. An AI model known as a “transformer” is used to learn the relationships among the data. Another language model generates explanations for the predictions.

The model was developed over 3 years using research in psychology, neuroscience, and genetics. The team initially studied thousands of people through psychological tests and various behavioral scenarios. Government and biobank data covering more than 2 million people was then added to the system. The customer-related data grew to cover approximately 30 million people. 42AI says it does not need personally identifiable information. Nevertheless, the system creates a separate behavioral profile for each person based on interconnected actions.

42AI claims that its market transaction predictions are 98% accurate. However, this does not mean that the system correctly predicts every transaction. The figure means that the overall trading flow falls within the predicted range. Its F1 score for transaction direction and volume ranges from 0.8 to 0.85. This metric reflects the balance between correct results and false signals. Factors the system cannot see, such as advice from a friend or colleague, may reduce the accuracy of its predictions.

The company currently works with 10 major brokers. Brokers use the platform to manage cash flow, margin calls, and hedging transactions. Customers pay a fixed fee or a certain percentage of the additional value generated by the platform. 42AI says its revenue is in the 7 figures. According to the company, one broker generated approximately $30 million (51 million manats) in additional value by managing risk more flexibly. The $62.5 million in funding came from the founders and Infinity Group over 3 years. In other words, not all of the funds were provided by external investors.

After the financial sector, 42AI is exploring adapting the model for gaming, fraud prevention, and cybersecurity. In the long term, the system will be used to help robots and AI agents better understand people. To achieve this, the model must be separately trained on the rules and conditions of each field. On the other hand, there is a risk that brokers could use customer behavior data for their own benefit. Although the company points to regulations, oversight, and an internal ethics committee, it has not disclosed a specific mechanism for preventing misuse.