AI Startups Developing World Models Maintain Strict Secrecy Over Commercial Plans
The emerging field of artificial intelligence known as "world models" has attracted significant funding and industry attention, yet the leading companies in the sector remain highly secretive about their specific commercial plans. At the recent All In conference, industry leaders and suppliers discussed the potential of the technology alongside the persistent lack of transparency regarding what these startups are actually building.
World models are designed to automate spatial intelligence, offering applications that could span robotics, interactive video, and advanced self-driving systems. Despite the substantial capital and public interest surrounding the technology, prominent players in the space—such as Yann LeCun’s AMI Labs and Fei-Fei Li’s World Labs—currently rank low on the commercialization scale, focusing instead on foundational research and development.
Unclear Timelines and Product Plans
During a panel discussion at the conference, Michael Rabbat, a co-founder of AMI Labs and the company’s Vice President of World Models, declined to share specific details about the startup's current projects. "We’ll talk about it when we’re ready to talk about it," Rabbat stated. He later clarified via email that the company, which is less than a year old, remains in a "research and building phase" and is not publicly discussing product plans or timelines.
This level of secrecy is characteristic of the broader world-modeling sector. Even World Labs, which features one of the most developed products in the space called Marble, has primarily used its technology for demonstration purposes. Marble's demos include media creation, generating explorable environments for video games, CGI effects, and potential robotics applications, rather than a singular, clearly defined commercial product.
Suppliers Kept in the Dark
The lack of disclosure extends beyond the public to the very suppliers providing data to these AI developers. Alex de Vigan, the CEO of data supplier Physicl, expressed frustration regarding the lack of communication from his clients in the world model sector. Speaking on the sidelines of the conference, de Vigan noted that while he knows Physicl's data is being utilized, he remains unaware of the final products being developed.
"I wish they would tell us more," de Vigan said. "We could build more useful data if we knew what they were working on." This disconnect highlights how closely guarded these development pipelines are, even from essential business partners.
Versatile Applications and Competitive Risks
The versatility of world models contributes to the mystery surrounding their ultimate commercial applications. At their simplest, world models function as navigable maps, similar to the AI systems that power self-driving vehicles. However, the same underlying technology could also enable humanoid robots to perform physical tasks like carrying boxes, or convert short video clips into fully explorable virtual environments.
AMI Labs has already explored diverse sectors, including manufacturing, biomedicine, robotics, and AI software for medical professionals through its partnership with Nabia. However, the company has not indicated which of these areas will become its primary commercial focus.
Industry analysts suggest there is little pressure for these startups to narrow their focus as long as venture capital funding remains readily available. Furthermore, maintaining secrecy helps prevent attracting immediate competition. If a company like AMI Labs were to announce a specific product—such as a humanoid robot component or a Hollywood-grade rendering system—it would likely draw rapid competitive responses from rival world model startups, neolabs, and established AI giants like OpenAI and Anthropic. By remaining quiet, these companies hope to delay intense market competition for as long as possible.
