The Dark Forest of AI: Why World Model Labs Are Going Silent

At the All In conference, the divide between ambition and execution was palpable. Michael Rabbatt, co-founder of AMI Labs, deflected questions regarding product timelines, citing a research-heavy phase. This reticence isn't an anomaly; it is a defensive posture. Even Alex de Vigan, CEO of data supplier Physicl, admits he provides the building blocks for these models without knowing their final destination, noting that better transparency would yield more tailored data outputs.

The versatility of world models—which could theoretically power anything from humanoid robotics and complex self-driving systems to Hollywood-grade CGI—is both their greatest strength and their biggest liability. By keeping their specific goals hidden, companies avoid signaling their intent to rivals. In an environment where fundraising remains relatively easy, there is little incentive to commit to a single vertical. Publicly declaring a breakthrough in manufacturing or robotics would immediately invite competition from deep-pocketed giants like OpenAI and Anthropic. For these labs, the current market dynamics function like a Dark Forest: the safest path is to remain invisible until the moment of deployment.