How Data Floats Away

Here, Take My DNA

Think back to the early days of consumer DNA testing. Millions of people sent their genetic material to services like 23andMe in exchange for novelty insights. What most didn’t realize was that they were giving away their most personal sovereign asset, their DNA. Years later, that data was sold, shared, and repurposed in ways customers never envisioned.

Data hovering works the same way. Every time proprietary documents, customer interactions, source code, or designs get fed into a public AI service, they can be analyzed, abstracted, and folded into models that others, including competitors, can benefit from.

Once you give away your DNA, you give away sovereignty over the most personal dataset imaginable. That’s the same tradeoff some enterprises make with their intellectual property today.

You can’t put the genie back in the bottle.  Once it’s gone, it’s gone.

Data Is Leverage

For years, we’ve heard the clichés: “data is the new oil,” “data is digital gold,” “data is the fuel for AI.” These metaphors sound catchy and might get clicks, but they do not hold up. Oil, gold, and fuel all share two traits: they are finite, and they are consumed when used. 

Data does not work that way.

Oil has value because it is scarce and must be drilled, extracted, refined, and then burned. Once consumed, it is gone. Data is the opposite. It can be reused infinitely without depletion. Its value does not come from being extracted in bulk but from how uniquely it reflects your customers, your operations, or your market.

Data is not digital gold. Gold is prized because it is rare and universally interchangeable. One ounce of gold has the same value as any other ounce of gold. Data is not interchangeable. Not all data is created equal, and that could not be truer here. A million rows of generic data are worthless if they are the same as what everyone else has. The value of data lies in exclusivity, not in how much of it you stockpile.

Data is also not the fuel for AI. Fuel powers an engine in a one-time transaction. You pour it in, it burns, and it’s gone. Data does not work that way. If anything, data is more like solar power. It can be reused, recombined, and recontextualized again and again. The same dataset can power multiple models and insights over time. Treating it like fuel ignores its renewable potential.

The World Economic Forum has argued that data is not oil at all, because it is not finite or rivalrous. Its value depends entirely on how unique and contextual it is. McKinsey has reinforced this view, showing that differentiated outcomes come from proprietary data tied to your customers, products, and operations, not from generalized bulk datasets.

Your data is your leverage. It's what makes your customer insights, your product designs, your financial models, and your operational playbooks unique. When that leverage hovers into someone else’s pipeline, you have traded away strategic advantage for temporary convenience.

Is it worth it?

How Data Floats Away

Several forces are accelerating the problem of data hovering today.

Public AI platforms are one such force. Many generative AI tools still use user inputs for training unless you purchase enterprise-grade, private deployments. The risks of training on copyrighted or proprietary material were highlighted by the New York Times v. OpenAI lawsuit.

Jurisdictional overreach is another challenge. The U.S. CLOUD Act, for example, lets American law enforcement compel U.S. providers to produce data they control, regardless of where it is stored. That collides with European privacy safeguards, which is why the EU’s highest court invalidated the EU–U.S. Privacy Shield in 2020.

Free copilots, chatbots, and plug-and-play AI tools make it easy for employees to paste sensitive information into public systems without realizing the long-term risks associated with data hovering. Use cases around governance are why guidance like the National Institute of Standards and Technology’s (NIST) AI Risk Management Framework exists, making it crystal clear that without explicit controls for data, models, and risk — AI cannot be trusted. 

Weak governance is like building systems without disaster recovery.  Everything looks great until it fails, and by that time it is too late.

The Cost of Avoidance

Ignoring data hovering is no different from ignoring backup or disaster recovery. You may save money and time in the short term, but when the failure happens, the cost of your compliance armor cracking is catastrophic. Just like failing to meet recovery objectives can break regulatory requirements, letting data drift into unauthorized jurisdictions exposes you to fines and lawsuits. Similarly, organizations without disaster recovery are often forced to pay ransom to regain access; with data hovering, you end up paying to use AI capabilities built on the back of your own surrendered data.

Another of the largest disadvantages of avoidance is the evaporation of your competitive edge. Where lost backups mean permanent data loss, what was once your unique advantage becomes absorbed into the same models your competitors rent.

Your brand and loyalty can also take a hit. Customers lose faith in companies that cannot protect or recover their systems after an outage. The same is true when they discover their data was quietly fed into a public AI model. Trust, once broken, is hard to win back. Just like customers may leave a company that cannot recover its systems, they may also leave a business that cannot protect their data from exploitation.

Avoidance is not neutral. It is surrender.

The Counterattack: Sovereign AI by Design

We all know disaster recovery cannot be bolted on after the fact. It has to be built into the architecture. Designing sovereign AI works the same way.

• Keep sensitive workloads inside your own virtual private cloud or datacenter.
• Govern model training with explicit consent and auditable controls.
• Treat data as leverage, not exhaust. Guard it as tightly as you guard your people and your capital.

Sovereignty does not happen by accident. It is engineered, just like resilient systems are engineered with backup and recovery from day one.

Data Anchoring

The era of data hovering is already here. Every day your data either stays anchored under your control or drifts into someone else’s empire.

You have two choices: anchor your data, designing systems that defend your leverage the way backup and disaster recovery defend your continuity; or, let it drift, handing over your advantage and renting it back later.

Because once your data floats away, it rarely finds its way home.

Data without anchoring is data without sovereignty. Ground it. Guard it. Own it.