A panel discussion at the All In conference this week highlighted the growing secrecy surrounding world model research, a field focused on teaching artificial intelligence systems to understand three-dimensional space. The session featured Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs, both of which have attracted significant investment and public interest despite remaining in early development stages.

World models aim to create AI systems capable of perceiving and interacting with virtual or physical environments, with potential applications ranging from robotics and autonomous vehicles to immersive media and digital simulation. However, companies in the space have been reluctant to disclose specific product roadmaps or commercial timelines. Michael Rabbat, co-founder and VP of World Models at AMI Labs, stated during the panel that the organization remains in a research and development phase and will not discuss product plans until they are ready for public announcement.

Alex de Vigan, CEO of Physicl, a data infrastructure provider serving several world model companies, confirmed that his firm supplies information used in these systems but lacks visibility into their final use cases. "I wish they would tell us more," de Vigan said, "so we could build more targeted and useful data."

While some demonstrations, such as those from World Labs’ Marble platform, showcase capabilities in generating interactive environments and visual effects, the broader industry has yet to reveal concrete pathways to monetization. The versatility of world models — capable of supporting applications in manufacturing, biomedicine, and AI-assisted healthcare — means companies may pursue multiple directions before settling on a focus.

The lack of transparency reflects broader strategic considerations. Public disclosure of a working prototype could attract competition from both emerging startups and established tech firms, including OpenAI and Anthropic. As a result, some firms appear to be delaying visibility to maintain a competitive advantage during the research phase.

This approach mirrors broader patterns in high-stakes technological development, where early secrecy may reduce the risk of premature imitation. Nonetheless, the pace of progress and eventual market entry remain uncertain, with no confirmed timelines or commercial launches publicly announced.

The article was published on September 18, 2026. Russell Brandom has covered technology policy and emerging tech since 2012, previously working at The Verge, Rest of World, Wired, and MIT’s Technology Review. He can be reached at russell.brandom@techcrunch.com or via Signal at 412-401-5489.