Humans& raises $480 million to build social intelligence AI

Humans&, a three-month-old startup founded by alumni from Anthropic, Meta, OpenAI, xAI, and Google DeepMind, has raised a $480 million seed round. The company is building a new foundation model architecture—defined as a large-scale AI model architecture—designed for social intelligence rather than just information retrieval. The startup does not have a product yet, but its founders aim to address coordination challenges as organizations transition from simple chatbots to active AI agents.

To achieve this, the startup will train its model using long-horizon reinforcement learning (RL)—a reinforcement learning technique to train models to plan and act over time—and multi-agent RL, which is reinforcement learning for environments with multiple AIs and/or humans. Yuchen He, a Humans& co-founder and former OpenAI researcher, noted that this training approach will involve more humans and AIs interacting and collaborating together. Co-founder Andi Peng, a former Anthropic employee, added that the industry is moving past a first paradigm of scaling specialized question-answering models and entering a second wave where average users are trying to figure out how to apply these tools. Rather than plugging into existing applications, Humans& is building a system centered on communication and collaboration. Eric Zelikman, the co-founder and CEO of Humans& and a former xAI researcher, explained that the company is building a product and model focused on communication and collaboration to help people work together more effectively.

The startup enters a crowded space for AI-driven collaboration. For context, peer startup Granola raised $43 million at a $250 million valuation. Industry figures like Reid Hoffman have also argued that coordination is the next major frontier for AI, representing the real leverage for how teams share knowledge. Humans& aims to position its model as the connective tissue across organizations, whether they are a family or a 10,000-person business.

Despite the high-profile interest, the startup is not interested in being acquired. Eric Zelikman, co-founder and CEO of humans& and former xAI researcher, stated, “We believe this is going to be a generational company, and we think that this has the potential to fundamentally change the future of how we interact with these models.”

Why it matters

Humans& is attempting to build a new foundation model architecture designed for social intelligence and coordination, aiming to act as a “central nervous system” for the human-plus-AI economy.