Insurance technology deals are often evaluated around speed, functionality and implementation risk. The quieter question is what the carrier actually owns after the technology starts working. If a platform improves underwriting, claims, property inspection or risk monitoring, but the underlying data remains trapped inside a vendor relationship, the insurer may be renting one of the most valuable assets it thinks it is building.
That issue is becoming more important as insurers rely on external data, AI tools and subscription-based platforms for core business functions. The National Association of Insurance Commissioners has formed a Third-Party Data and Models working group to develop a framework for regulators evaluating third-party AI data and models, and its AI systems evaluation tool is being piloted by 12 states in 2026. Legal dealmaking is moving the same way: AI-specific contract provisions increasingly address ownership, inputs, governance, third-party services and performance claims.
For Rob Galbraith, CEO of Forestview Insights, the risk is practical before it is theoretical. At a prior carrier, his team was evaluating aerial imagery providers to obtain property images and repeated snapshots over time.
“We’re trying to get good quality aerial imagery of properties and then be able to get those repeated snapshots over time and see if we detect any changes,” Galbraith said, citing examples such as new structures on a property or roof deterioration.
Then came the contract question that should sit much higher in insurance technology decisions: “Is this licensed? If we go to another vendor, now we’ve lost access to all of this, right? Or do we own these images?”
The answer mattered because property data becomes more valuable as it accumulates. One image can support a point-in-time underwriting or claims decision. A sequence of images can show deterioration, repair, mitigation, additions, vegetation change or a discrepancy between what was represented and what exists. That history can support underwriting, claims handling, catastrophe exposure management and conversations with reinsurers or regulators.
Galbraith said the selected vendor’s position was a deciding factor: images the carrier purchased remained the carrier’s to keep, even though future updates would stop if the relationship ended. That distinction separates a technology subscription from a data asset.
“So many of these subscriptions as a service has been a popular paradigm with a lot of different technological applications,” Galbraith said. “But it can lead to some of the data ownership part.”
For insurance executives, that is the hidden lock-in. A carrier may believe it is building institutional knowledge about its book, only to discover that switching systems also means losing historical records, derived attributes, image archives, timestamps, metadata or model outputs. The cost is not only contractual. It is operational memory.
Data ownership is not limited to who possesses files at contract termination. It also affects whether the insurer can understand how data was sourced, validated and used in models that support business decisions.
“How’s your model trained?” Galbraith said. “Is it all garbage in, garbage out because you just sucked in everything from the Internet, or do you actually have a validation process?”
That question matters in a regulated industry. Even when an insurer uses a third-party tool, the carrier remains exposed to questions about the quality of the data behind pricing, underwriting, claims or fraud signals. If the data source cannot be explained, exported or audited, the carrier has accepted a risk it may not be able to defend.
The strategic question is therefore not only whether a tool works. It is whether it leaves the insurer with a durable asset. Executives should know which data is purchased, which is licensed, which outputs the carrier owns, what happens at termination, whether historical archives can be exported in usable formats and whether the carrier’s operational data can train systems used elsewhere.
That does not mean rejecting subscription-based technology. It means treating data rights as part of transformation architecture, alongside workflow design, cybersecurity and regulatory governance. In property, claims, underwriting and AI-enabled analytics, the value often compounds over years. Losing the record can be as damaging as losing the system.
For carriers buying technology in 2026, data ownership should not be buried in procurement. It belongs in the business case. The insurers that understand what they own will be able to change vendors, prove decisions and keep learning from their own books. Those that do not may find that the transformation they paid for lives somewhere else.
