We live in an age rich in data collection. Computing power and data storage are astoundingly cheap, and machines bristle with sensors, from the internal gyroscopes that let cellphones count steps to the massive electro-optical cameras strapped on drones recording hundreds of hours of battlefield observations. To turn that raw data into processed, useful information, the Pentagon wants to turn to the technology giants of Silicon Valley. The next hurdle: keeping that data from being siloed away into a single company’s proprietary tool.

“We don’t necessarily have to own [the data],” according to Michael Brown, managing director at Defense Innovation Unit. He spoke last month at a roundtable of tech executives and Pentagon leaders held in California.

The Pentagon itself doesn’t have to own all the data churned to power artificial intelligence, but it does want to make sure that data can be shared across multiple contractors. Trae Stephens, co-founder and chairman of the Peter Thiel-backed Anduril Industries and a partner in the Peter Thiel-backed Founders Fund, points to the Thiel-backed defense contractor Palantir as a example of how to correctly handle data.

“So for a company like Palantir, for example, there’s never any assumption that they own the data,” Stephens said. “The data is the clients, the software is running on the client’s own servers and in the usual case.”

Palantir has, from early on, received funding from and contracted to the intelligence community and the Department of Defense, so it should perhaps come as little surprise the company built its data handling in a way that reinforces Pentagon priorities.

“I think that, broadly speaking, the better technology companies probably have an open API [application programming interface] mentality about data sharing as it is,” Stephens said. “It’s actually very old school to believe in, you know, not having API access to data.”

Besides being old-fashioned, why might a company want to keep a tight hold on the data it’s been given? Well, because holding data itself is valuable, even and perhaps especially if others could use that data better.

“It’s that if they make that data accessible to everyone, they lose a significant advantage over keeping the department locked into using their system, and they’re worried about being able to recoup their investment,” said Josh Marcuse, executive director on the Pentagon’s Defense Innovation Board.

Marcuse sees fights over data usage rights as a proxy for a trickier issue: vendor lock. Contracts may be written in such a way that lets the Pentagon buy the data, but then the vendor may filter that data through a proprietary interface, making it difficult to use the same data across multiple tools made by multiple providers. Instead, the Pentagon is stuck working with the company that locked in the first contract.

Processing the data, and then providing it in a locked, proprietary format may deliver the immediate product solicited by the Pentagon, but the downstream effect becomes one where the Pentagon simply returns to the company it previously hired, rather than trying to extract the data, reformat it, and offer it to a different company that could provide a better and/or cheaper service.

“It makes it that much harder for the department to improve the quality of the products and services that it provides the war fighter,” Marcuse said. “The department deserves to have good, new stuff. The way the contracts are written, it gets really hard to do that.”

Finding a way to navigate those competing impulses — that of a buyer who wants ease of sharing data and a contractor looking to build a durable income stream from its data analysis and processing — likely means discovering a new business model, suggested Marcuse. That makes the overall issue less about data rights and more about finding a way to write the contracts so that the contractor feels it has a viable data model without forcing vendor lock, and so the Pentagon can receive the analysis it needs.

“I do think it’s changing,” said Siobhan McFeeney, vice president of transformation at technology firm Pivotal. “I started to see some of the companies start to open their minds and [say]: ‘OK, I guess we’ve got to figure out how to build software differently because that thing we had maybe isn’t quite as fantastic as we thought it was and time’s running out.’ I imagine the next 12 months will be very interesting.”

Even if a change in business model comes to pass, skepticism is warranted regarding the openness of this data. That openness will likely mean, at most, moving data from single-contractor silos and walled gardens to the slightly larger confines of a five-sided building in Arlington, Virginia. The goal isn’t the free flow of data outside the Pentagon, so much as a free flow of data within it.