Humans& AI Coordination Models

What happened

Humans&, a three-month-old startup founded by alumni from Anthropic, Meta, OpenAI, xAI, and Google DeepMind, secured a $480 million seed round, valuing it at $4.48 billion. The company is developing next-generation foundation models specifically for coordination and collaboration, not chat. These models utilise long-horizon and multi-agent reinforcement learning to manage group decisions, memory, and multi-user workflows, aiming to establish an AI messaging substrate for team coordination.

Why it matters

The introduction of AI foundation models explicitly designed for complex multi-agent collaboration, without a publicly available product, creates a new dependency and oversight burden for platform operators and IT security. Evaluating these systems requires increased due diligence to understand their interaction dynamics, memory capabilities, and potential failure modes in multi-user workflows. This shifts the control focus from individual AI agent behaviour to the integrity and security of the entire collaborative AI layer, potentially introducing a visibility gap in system-level accountability and data governance.