Israeli startup 42AI says its world model can predict human behavior

Tel Aviv startup 42AI emerged from stealth on November 2 with a "behavioral world model" it says is already running inside 10 brokerages and reading close to 30 million people, backed by $62.5 million built up over three years from the founders and Infinity Group, the vehicle of chairman Amir Gal-Or. CEO Amit Gal-Or told TechCrunch the model claims 98% accuracy predicting market trades, a figure he said measures whether trading flows land inside a predicted range rather than whether each individual trade is called correctly, and he said "the model never knows who the person is." The funding came without an external venture round, contrasting with World Labs' $1 billion raise and Yann LeCun's AMI Labs at roughly €890 million.

Ten brokerages are already running 42AI's model. The Tel Aviv startup says it can explain why people act, and it has $62.5 million behind the claim. Close to 30 million people are being read by software they've never heard of. Co-founder and CEO Amit Gal-Or says 42AI's model is already running inside 10 brokerages, parsing their trading patterns. The Tel Aviv startup came out of stealth on November 2 with what sounds almost too clean a pitch: feed it behavioral data, and it will tell you not just what someone might do next, but why. According to TechCrunch's Ram Iyer, who first reported the launch, 42AI calls its product a "behavioral world model". That's the detail worth sitting with. Most AI products selling "predict your customer" promises are language models wearing a costume, prompted or fine-tuned to role-play as a 35-year-old suburban shopper or a nervous day trader. 42AI built something different from scratch, and the distinction is the whole business. Gal-Or told TechCrunch the model treats behavior as probabilistic rather than fixed, arguing that people carry underlying psychological tendencies that persist even as their circumstances change. In practice, that means the system is trying to model a person's internal motivations and constraints rather than just extrapolating from what they did last time. It's a world model in the technical sense now associated with Google DeepMind's Genie, Meta's V-JEPA, and Fei-Fei Li's World Labs. That means a system trained to predict the next state of something, not just the next word. 42AI's bet is that the "something" worth predicting is us. The brokerage use case is where 42AI has real numbers to show. The company claims 98% accuracy predicting market trades. But Gal-Or was careful with TechCrunch about what that figure actually measures: whether trading flows land inside a predicted range, not whether the model calls each individual trade correctly. That's an honest caveat. It matters, because "98% accurate" without it is the kind of number that gets stripped of context the moment it hits a pitch deck. Nace.AI's Drex v2.0 bets agents don't need words to make a decision https://startupfortune.com/naceais-drex-v20-bets-agents-dont-need-words-to-make-a-decision/ Nace.AI's Dos Bakytzhan built the Palo Alto startup on $26.5 million in funding from Jerry Yang and an Intel executive, and Drex v2.0 now edges rival Jev 1.13.0 by 3.3 points on the Decision Index benchmark. - AI agents decision making without language models https://startupfortune.com/naceais-drex-v20-bets-agents-dont-need-words-to-make-a-decision/ - Drex v2.0 probability scoring for agent decisions https://startupfortune.com/naceais-drex-v20-bets-agents-dont-need-words-to-make-a-decision/ Then there's the privacy question that any "predicts human behavior" headline invites. Gal-Or's answer to TechCrunch was blunt: "the model never knows who the person is." Does that hold up against the kind of re-identification risk that has dogged plenty of anonymized datasets before it? Outside researchers will want to test the claim, not just take it on faith. Money that didn't come from a VC pitch deck Here's what separates 42AI's funding story from the rest of the world-model gold rush: the $62.5 million wasn't raised in a splashy external round. It came from the founders themselves and from Infinity Group over three years of building in stealth, per TechCrunch's reporting. Infinity Group was founded by Amir Gal-Or, who is 42AI's chairman. So the money has more of a family resemblance to a founder-backed vehicle than to a traditional seed raise with a term sheet from a name-brand fund. That's a very different posture from the rest of the field. Yann LeCun's AMI Labs raised roughly €890 million. World Labs closed a $1 billion round with Nvidia and AMD at the table. Mirror Particle, a rival chasing the same human-behavior niche, was still closing its first venture round as of October while competing in TechCrunch's Startup Battlefield. Simile, another human-behavior-prediction startup, raised $100 million from Index Ventures and Bain Capital Ventures. Against that backdrop, 42AI choosing patient, insider capital over a competitive raise reads less like modesty. It looks more like a company that wanted three years of runway without a board clock running. Where this goes next should interest founders and investors more than the brokerage pilots. Gal-Or told TechCrunch the company is looking at gaming, fraud prevention and cybersecurity as its next markets. The long-term ambition is helping robots and AI agents understand people well enough to act around them. Fraud detection and cybersecurity are a believable next step. A model trained on probabilistic human tendencies should be good at behavioral anomaly detection. It's also a market that already pays well for marginal gains in accuracy. The harder sell is anything touching advertising, hiring, insurance or political targeting, the four places a "predicts what you'll do" model turns from a trading tool into a civil liberties question. 42AI hasn't announced products in any of those categories yet. But the model doesn't need to be deployed there for the debate to start. A system that claims to infer motivation from behavioral data is already going to get test-driven the moment the API is open, and the company that built it says it works. Everyone from ad-tech firms to political consultants will want a go. Whether 42AI lets them in is a decision the company hasn't had to make publicly yet. It will. Also read: DeepSeek Turns Creator Mode Into a Land Grab for Its Harness https://startupfortune.com/deepseek-turns-creator-mode-into-a-land-grab-for-its-harness/ • North Korea Uses AI Faceswaps and Voice Cloning to Land Fake IT Jobs https://startupfortune.com/north-korea-uses-ai-faceswaps-and-voice-cloning-to-land-fake-it-jobs/ • Open-Source Fix Slashes DeepSeek V4.1 Startup Time on Gaming GPUs https://startupfortune.com/open-source-fix-slashes-deepseek-v41-startup-time-on-gaming-gpus/ This article is posted in AI News https://startupfortune.com/category/ai/ , check it out for more related stories. Join the discussion Open in the community → https://startupfortune.com/community/ Almost there. Sign in and your reply posts straight away.