42AI Exits Stealth With $62.5M for Behavioral AI Models

42AI has emerged from stealth with $62.5 million in seed funding to develop AI models that predict human decisions, TechCrunch reported. The Israeli-American startup built its behavioral technology over roughly three years, drawing on psychology, neuroscience and behavioral modeling, according to 42AI.

The money came from 42AI’s founders and Infinity Group over approximately three years, rather than a single entirely external financing round, TechCrunch reported. Co-founder and CEO Amit Gal-Or and co-founder and CTO Ehud Ezra lead the company.

What does 42AI predict?

42AI initially focuses on financial services, where its behavioral world model estimates how individuals respond to different circumstances, the company said. The technology helps brokerages anticipate customer trading activity and manage financial exposure, according to 42AI.

42AI reports that 10 large financial brokerages use its technology to support liquidity management, margin requirements and hedging decisions. Its initial commercial focus centers on customer behavior inside those institutions, according to the company.

How does the model work?

Amit Gal-Or, 42AI’s co-founder and CEO, said: “Behavior is not deterministic; it’s constantly probabilistic.”

Ehud Ezra, 42AI’s co-founder and CTO, said: “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 system analyzes patterns across four types of customer interaction, 42AI said:

• Customer onboarding
• Support calls
• Fund transfers
• Investment transactions

42AI converts those observations into mathematical representations of behavioral tendencies. A transformer-based AI model processes the representations and calculates probabilities for potential future decisions under different conditions, the company said.

Rather than rely exclusively on historical transaction patterns, 42AI aims to model underlying behavioral characteristics that may influence how people respond as circumstances change, according to the company.

42AI initially studied thousands of individuals through psychological assessments and behavioral scenarios. It later incorporated government and biobank datasets covering more than two million people, the company said. Its customer datasets now encompass information associated with approximately 30 million individuals, according to 42AI.

The company also consulted specialists in psychology and post-traumatic stress disorder, along with Nobel Prize-winning chemist Dan Shechtman, according to 42AI.

The company’s current data scale reaches approximately 30 million individuals, according to 42AI.

What do the performance numbers show?

42AI reports a 98% accuracy figure for whether trading flows fall within predicted ranges, not for correctly forecasting every individual trade, according to the company.

For predictions involving trading direction and volume, 42AI reports F1 scores ranging from approximately 0.80 to 0.85.

In 42AI’s reporting, the 98% figure measures predicted ranges, not correct calls on every individual trade.

How does 42AI make money?

Customers pay fixed fees or arrangements based on the financial value generated by the platform, 42AI said.

One brokerage generated approximately $30 million in additional value through dynamic hedging supported by the technology, the company reported.

42AI generates annual revenue exceeding $1 million, though it has not provided a more precise figure, according to the company.

The platform has crossed $1 million in annual revenue, according to 42AI, though its precise figure remains undisclosed.

What comes next?

42AI is exploring applications in gaming, fraud prevention and cybersecurity, as well as potential interactions between autonomous AI agents and human users, the company said.

It also plans to pursue additional financing as it expands development and commercial applications of its behavioral AI technology, according to 42AI.

TechCrunch’s report raised questions about behavioral profiling, privacy and the possibility that financial institutions could use predictions of customer behavior for their own advantage.

42AI says it uses regulatory compliance processes, internal oversight and an ethics committee to address those concerns. Questions remain about how customers may ultimately use the system’s predictive capabilities, TechCrunch reported.

For 42AI, the next test is whether its platform can scale beyond 10 large brokerages as it adds gaming, fraud-prevention and cybersecurity applications and seeks more capital, according to TechCrunch.