Who Owns the Spatial Data Collected by Industrial Robots?

When we think about industrial robots, we tend to think about what they can do. How accurately can they function? How safely can they work around people? And how much labour can they relieve us of? But there’s another side to robotics, and it’s just as important: the information these machines gather while going about their business. 

Modern robots utilise sensors and cameras to construct precise images of their environment. This information allows them to navigate, recognise objects and respond to changes in their surroundings. But it can also be saved, analysed and used for purposes other than the robot's primary duty. 

A noteworthy example is Hesai's October 2026 debut of its Kosmo 3D World Data Engine. The platform aims to turn captured real-world environments into reusable digital resources for applications like robotics training and simulation. So who controls that information when a robot is gathering precise data in a factory or warehouse?

And what happens when that information becomes more valuable than the original sensor readings?

What Spatial Data Do Industrial Robots Actually Collect?

Before we discuss industrial robot data ownership, we need to get a grip on what we’re talking about. Because the information a robot receives may not be the information it ultimately uses, stores or distributes. Robots employ cameras, depth sensors and LiDAR to interpret the world around them, depending on their design and purpose. 

LiDAR stands for Light Detection and Ranging and uses laser measurements to work out distances. These measurements can form a point cloud, a collection of points that denote the locations of objects and surfaces in three-dimensional space. Imagine it as a rough 3D sketch of a room. It might reveal the walls, machines and barriers without resembling a snapshot. 

Cameras can provide visual detail, and depth sensors can measure the distance of various objects. Software can then process these signals to produce maps or even more detailed 3D models. For example, a mobile robot might use a map to work out where it is (known as localisation) and plot a course around the factory floor. 

Not all robots remember everything they sense. Some information is processed promptly and then thrown away. Other data may be stored to update maps, analyse faults or improve performance. There’s also a difference between data obtained in normal operation and data intentionally prepared for simulation or robot training. 

One can help a machine do the task at hand, while the other can help create or test future capabilities. The significance of that difference becomes apparent when the same information can serve many purposes.

Why Robot-Generated Spatial Data Is Becoming More Valuable

A robot might simply need to recognise a shelf to move around it safely. But a detailed digital representation of the warehouse could be useful long after that particular journey is over. It might enable engineers to test various layouts, simulate robot movements or prepare robots to work in unfamiliar environments. 

And since the environment already exists in digital form, some of that work may be done without disrupting the physical operation. This is part of what makes spatial data increasingly valuable for physical AI, where intelligent systems need to perceive and interact with the real world.

NVIDIA's Isaac Sim documentation, for example, describes capabilities for creating synthetic data in simulated environments. These environments allow developers to build training and testing scenarios without needing to physically reproduce every circumstance. Kosmo works in a similar way, taking real-world environments and turning them into digital resources. 

Those assets can then feed additional simulation and training processes. The fact that a supplier can technically use a dataset for training doesn’t guarantee they actually do. Different platforms have different capabilities, and actual data use depends on the system and its agreements. But it does raise an interesting commercial question.

If information collected for one purpose can also be used to produce additional products or services, who is entitled to benefit from that additional use?

Who Actually Owns the Data Collected by Industrial Robots?

There isn't a single, standard answer for who owns industrial robot data. It depends on the kind of information, agreements between parties and applicable legislation. Purchasing a robot doesn’t mean that an enterprise owns all the digital resources associated with it. 

Similarly, a manufacturer that has access to sensor data doesn't automatically have carte blanche to utilise that data however it chooses. A robotics deployment can include the firm that operates the equipment, the robot manufacturer, a systems integrator and a separate platform provider. They may each have separate responsibilities and rights. 

The distinction between raw and processed information is particularly relevant here. The European Commission’s guidance on the EU Data Act helps explain why. The law has applied since 12 September 2025 and gives users rights to access certain data generated through their use of qualifying connected products, including industrial machinery and robots. 

But these rights aren't absolute. The applicable provisions usually cover readily available raw and pre-processed data, including associated metadata. Inferred or derived information, such as highly enriched datasets, isn't included in that particular scope. So an enterprise may have rights to particular sensor readings, but not automatically to every processed 3D model generated from that information. 

The Act also contains provisions concerning personal data, trade secrets and security. Its application depends on the circumstances, and it can't replace clear contractual arrangements. For enterprise buyers, the question isn't only ownership. They need to know what they can access, what they can do with it and whether they have rights to information that results from subsequent processing. 

Having a copy of a dataset doesn’t mean you have authorisation to reuse or distribute it.

What Are the Security and Privacy Risks of Spatial Data?

Even if a precise 3D map doesn’t contain customer data or financial information, that doesn’t mean it has nothing sensitive to reveal. Think of what a person could learn from an accurate representation of a manufacturing plant. Depending on what's captured, it could display machinery locations, restricted areas, access routes or details of how the operation is organised. 

The concern isn’t only that robots acquire this data. It’s what could happen if comprehensive spatial records were accessed or shared beyond their intended purpose. Researchers are already looking for ways to limit unnecessary disclosure. 

A 2026 study, RoboShape: Information-Theoretic Point Cloud Representations for Privacy-Aware Robot Perception, evaluated whether robots could retain important information for recognising objects while exposing less about potentially sensitive parts of their environment. 

Across three indoor LiDAR datasets, the researchers reported retaining 98.7% of object-classification utility while reducing sensitive-attribute prediction performance by 39.3%. These are experimental results, not proof of a commercially viable privacy solution. 

But the discovery suggests an intriguing possibility: robots might not have to remember every detail their sensors can pick up. Enterprises need to consider what happens to information once it's collected as well. Some systems process data locally, while others may use cloud services or third-party platforms. 

The US National Institute of Standards and Technology's Guide to Operational Technology Security already provides guidance for managing security risks in systems that interact with physical environments. For spatial data, the practical issues include managing access, protecting transfers, reducing unnecessary retention and making sure information can be removed when it's no longer required.

What Should Enterprises Establish Before Deploying Industrial Robots?

By the time a robotics system is working across a factory, its data collection and processing arrangements may already be incorporated into routine operations. Changing them later could mean revisiting technical setups, supplier contracts or existing processes. This is why robotics data governance should be part of deployment planning and procurement. 

There are four areas worth getting clear on before committing to a system.

What data will the system create and retain?

First, ask the supplier what information the system actually produces.

• Is it capturing raw camera images, LiDAR point clouds or depth data?
• Does it produce maps, reconstructed environments or other processed datasets?
• And which of those outputs actually get saved?

It's also important to know how often information is gathered, where it's processed and how long it's kept. A robot may require detailed sensor data for safe navigation without needing to preserve a permanent record of everything it observes. Knowing the distinction helps companies decide what information is needed and what may be discarded.

Who can access and reuse the information?

Access rights are only half the story. Enterprises should also specify what each authorised party is allowed to do with the data. For example, a supplier may need access to some records for maintenance or troubleshooting purposes. But does that permission cover product development, training AI models or sharing information with another provider? 

Those applications aren't interchangeable. Contracts should specify who can access particular datasets, the purposes for which they can use them and whether further approval is required for secondary uses. Where appropriate, businesses should also consider whether data may be combined with information gathered from other customers.

How will the data be protected?

Security provisions should be appropriate to the information being gathered and the environments it depicts. This includes knowing whether processing happens on the robot, on local infrastructure or through external services. It also means identifying who manages access controls, encryption, monitoring and security incidents. 

Retention matters too. If a dataset is no longer needed, can it be deleted? And, subject to agreed retention requirements, does that deletion include copies stored by service providers and backups? The organisation, equipment manufacturer and any other supplier handling the information should have a clear understanding of these responsibilities.

Can the enterprise retrieve and reuse its data elsewhere?

The supplier may enable data export, but that doesn't ensure the exported data can be used in another system. Enterprises should determine which datasets are retrievable, in what formats and whether the files can be interpreted without proprietary software. For example, getting raw sensor information isn't the same as getting a usable 3D representation of the facility. 

Even if both are available, the rights associated with them may be different. It’s worth putting these capabilities through their paces before deployment, rather than relying on a broad guarantee of portability. Contracts should also specify what happens when the relationship ends. 

This includes access to retained data, rights to continue using processed assets and provisions for deleting information the supplier no longer requires.

Why Spatial Data Governance Needs to Start Before Deployment

There's a practical reason to settle these questions early. The initial configuration of a robotics system can determine what data will be available years down the line. A firm might initially need spatial data for navigation and obstacle detection. As its robotics programme grows, it may wish to use that data for simulation, testing or other purposes. 

But these possibilities may depend on decisions made long before anyone considered the additional uses. This isn't something procurement or IT can settle by themselves. Operations teams understand what the robots need to do, security teams understand what needs to be protected, and legal and procurement teams can define the rights and obligations. 

Bringing these perspectives together during deployment planning gives the organisation a better understanding of the information it's producing and the options it has. As physical AI advances, that knowledge may become increasingly useful.

Final Thoughts: Industrial Robot Data Rights Need to Be Clear From the Start

Industrial robots are gaining a better understanding of their surroundings. And as those capabilities develop, the information they acquire may be useful for far more than just accomplishing their immediate tasks. But owning the machine, having access to sensor information and having rights to reuse a detailed 3D model aren't necessarily the same thing. 

It's up to enterprises to understand the distinctions before making assumptions about what can be retained, shared or used in the future. The wider conversation around enterprise robotics is moving beyond capabilities to how businesses manage the technology behind them. 

EM360Tech continues to cover developments in physical AI and enterprise automation, including the decisions businesses need to make as these systems become part of their day-to-day operations.