Who Owns the Rain?

Photo by Valentin Müller on Unsplash

I never thought I’d have to write about rain ownership, but here we are.

Rain never used to have an ownership tag. Everyone depended on the traditional weather forecasts we all know about.

The ones paid for by the same agencies, and based on the same open data. Rain was one of the few genuine communal resources left. I’d put it up there with free parking and sunrises.

Unfortunately, the arrangement I just described is on its last legs.

AI is getting exceptionally good at predicting weather patterns, and the organizations responsible for building the most competent models have started whispering a question that’s making a lot of people squirm.

“If our algorithm produces a sharper forecast than the government’s, who actually owns that forecast?”

All these organizations are coming to the conclusion that they, in fact, own that data. This is opening up a can of worms and starting fights between insurers, tech companies, and the public agencies that used to run the show on their own.

The forecast used to be a public good

For the greater part of the twentieth century, the government was in charge of weather prediction. Agencies that include the U.S. National Oceanic and Atmospheric Administration and the European Centre for Medium-Range Weather Forecasts built physics-based models using publicly funded satellites and sensors, and then they gave away the output.

This practice was almost like a doctrine, and the World Meteorological Organization’s open data policy formalized that perception.

Weather forecasts were viewed as a shared foundation that everyone could build on.

Technically speaking, that foundation still stands (this statement is doing a lot of heavy lifting).

ECMWF now publishes its own AI-based forecasting system (AIFS) alongside the traditional models. It also continues to make sure a huge chunk of its data is freely available through what has been described as a deliberate transition from licensing to open access.

But the most efficient weather models today are not being built by public agencies.

Enter the private forecast

WindBorne, a Stanford-founded startup, is now making the bold claim that its AI model (WeatherMesh) outperforms weather forecasting systems built by the European government.

Apparently, this AI model uses sensor data fed into deep learning architectures that the company controls end to end, as reported by TechCrunch.

Swiss Re’s modeling subsidiary (Fathom) uses diffusion models, which is the same underlying technique behind AI image generators, to come up with tens of thousands of years of synthetic weather events for a hypothetical 2030 climate, filling in gaps that historical records do not cover, according to a Financial Times investigation reported by The Decoder.

None of the data generated from this goes into public archives. It goes to proprietary risk models that insurance companies are now using as their basis for pricing policies.

And of course, those insurers have every possible incentive to keep all that data hidden behind a login screen.

Most of us would agree that they were never really interested in public service. Access to this data just gives them a competitive edge, which means more money in their pockets.

Why insurers are suddenly obsessed with owning the data pipeline

Let’s look at the math behind all this. Globally, insured losses from natural disasters hit about $108 billion in 2025, with storms, floods, and wildfires responsible for most of that damage, according to figures from Munich Re.

In response, insurers are pouring bucketfuls of cash into AI. In the U.S., insurance technology spending is projected to rise by $173 billion in 2026, according to Forrester estimates.

In the midst of all this, executives are also nervous about what will happen if the public data underneath all this disappears.

NOAA data is still foundational to numerous climate and catastrophe models. Pullbacks in what is maintained and updated by the federal government are already sending waves of anxiety throughout the reinsurance world.

Enough that a private-sector initiative is reportedly being organized just to preserve those datasets independently, according to reporting from Insurance Business.

Simply put, the same companies that are racing to build proprietary weather intelligence are also worried about losing access to the free public data that was used to train their proprietary models.

The government’s side of the argument

Governments have their own reasons when it comes to why they have to be territorial. One of those reasons is safety, but it’s the least controversial one.

The United Nations’ Early Warnings for All initiative is attempting to make life-saving forecasts accessible to every country by 2027. As noble as this is, it entirely depends on data staying open instead of locked inside a handful of corporate models built to fuel the capitalist machine.

A great earthquake forecast is not that useful to a remote village in a country that cannot afford to pay the subscription.

There is another motive that happens to be more subtle and more territorial. Over time, weather and climate data have increasingly functioned as infrastructure instead of just information. From our history, we know that nations would rather control infrastructure than rent it from a handful of private companies that have their headquarters somewhere entirely different.

Some researchers are tracking the shift, and they have noted that some high-value data streams are now moving toward proprietary licenses and commercial infrastructure in a way that does not fit comfortably next to the decades of open-data practices that we are used to, according to a recent analysis of machine learning’s impact on forecasting practice.

So who legally owns the data?

Photo by Tingey Injury Law Firm on Unsplash

Nobody has settled on a legally satisfactory answer to who actually owns AI-generated weather data. The weather is not something that was meant to be owned by one person or entity in the first place.

We can’t apply copyright law here because it only protects creative expression, and definitely not facts about the atmosphere, which are much closer to simple facts.

But the model generating the forecast, the training data pipeline behind it, and the specific synthetic scenarios a company like Fathom comes up with are all things a lawyer could reasonably claim belong to somebody.

The ambiguity is why this is becoming a real fight.

WMO convened a conference of more than fifty experts from agencies, companies, including Google, IBM, Microsoft, and NVIDIA, and academic institutions. The agenda included discussions about how weather models should be governed.

The sentiment at the end of all this leaned heavily towards the idea that the global meteorological community should keep investing in shared, open tooling and benchmarks, according to WMO’s own summary of the event.

I think that’s just a diplomatic way of saying that people could not agree on where lines should be drawn.

What does this mean for the rest of us?

In the event that insurers and private forecasters come out on top (and I hope they do not), the most accurate weather predictions are going to be hidden behind a subscription.

The sad part is that in most cases, the people who need these predictions the most are the ones who are least able to afford the subscriptions.

If public agencies win, the consequences might be slower innovation, but everyone will still have access to the baseline information that helps one to decide how safe it is to stay in one’s house during an earthquake.

It is highly likely that neither side will have a clean win.

Public agencies still keep publishing baseline forecasts because letting people die in storms is a bad look for a democracy.

On the other side, insurers and tech companies keep building increasingly proprietary layers on top, which are sold to whoever has deep pockets.

Rain will still fall on everyone. Knowing when, where, and how hard it might fall might be something you will need to pay for in the near future.