Why World Model AI Companies Are Keeping Quiet

World models are attracting serious attention and funding in the AI industry, but some of the companies building them are being unusually careful about revealing what they plan to do with the technology.

At the recent All In conference, a panel discussion on world models offered a closer look at one of the more mysterious areas of AI. Two of the biggest names in the field are AMI Labs, founded by Yann LeCun, and World Labs, founded by Fei-Fei Li. Both companies have generated significant interest from investors, yet neither has clearly laid out a major commercial roadmap.

The technology itself has plenty of potential. World models are designed to help AI systems understand and interact with physical spaces, opening possible applications across robotics, interactive video, manufacturing, self-driving technology, and other areas that require spatial intelligence.

The question is where that technology will become a real business.

AMI Labs isn’t saying much yet

Michael Rabbat, a co-founder of AMI Labs and its VP of World Models, appeared on the panel and was asked about the company’s plans. His response was notably cautious.

Rabbat said the company would discuss its plans when it was ready. In an email, he offered a little more context, explaining that AMI Labs remains in a research and building phase and is not publicly discussing product plans or timelines.

AMI Labs is also less than a year old, giving the company a reasonable reason to keep its development work private.

But the secrecy isn’t limited to AMI.

World Labs shows the technology, but not the full roadmap

World Labs’ Marble currently appears to be one of the most developed products in the world-model space. Its demonstrations show how the technology can create media and generate explorable environments that could be used for video games or CGI effects.

Robotics is another possible application, although Marble currently appears to be focused more on demonstrating what world models can do than on showing a single commercial direction.

Even companies supplying data to the industry don’t always know what these AI labs are ultimately building.

Alex de Vigan, CEO of Physicl, a data supplier for the growing world-model industry, said he knows the data his company provides has been useful to these businesses. However, he doesn’t necessarily know how it is being used.

De Vigan said that if companies shared more about their goals, Physicl could potentially create more useful datasets for them.

One technology, many possible markets

A major reason for the uncertainty is the broad nature of world models.

At its simplest, a world model can function like a navigable representation of the physical world, similar in some ways to the AI systems used by self-driving vehicles. But the same underlying technology could potentially help a humanoid robot move objects around, create interactive digital environments from video, or support other forms of spatial computing.

AMI Labs has already explored areas including manufacturing, biomedicine, robotics, and AI software for doctors through its Nabia partnership.

That doesn’t mean the company intends to pursue all of these markets. Instead, it highlights just how many possible directions the technology could take.

Staying quiet could be part of the strategy

There is also a business reason for companies to remain vague.

As long as investors are willing to fund research and development, AI companies have less immediate pressure to commit publicly to one specific product or market. Revealing a breakthrough too early could also alert competitors to a potentially valuable opportunity.

If AMI Labs suddenly announced a major humanoid robotics system or a next-generation Hollywood rendering platform, other AI labs would likely take notice. Potential competitors could include other world-model startups, newer AI labs, as well as major companies such as OpenAI and Anthropic.

That creates an unusual situation: the same investment environment that gives companies the resources to develop technology privately is also giving potential competitors the money to pursue similar opportunities.

For now, keeping plans secret can give these companies more time to build before the market fully understands what they are developing.

The situation resembles what fans of science-fiction author Cixin Liu might call a “dark forest” scenario. When you don’t know who else is watching or what they are capable of, drawing attention to yourself may not be the smartest move.

For the world-model industry, that could mean more breakthroughs happening behind closed doors before companies are ready to reveal where the technology is actually headed.

This article was first published on September 18, 2026.

Also read : Mazama Energy Raises $135M to Advance Super-Hot Geothermal Power