Missing link: Power center Palantir – a software controls organizations

Contents

Few technology companies are as polarizing as Palantir Technologies. Founded in 2003, the US company has developed from a secretive start-up with CIA connections into a listed corporation that supplies governments and large corporations worldwide. Palantir promises to gain usable insights from the data of modern organizations. While supporters praise the software as a revolutionary tool for data-based decisions, critics warn of surveillance, a lack of transparency and vendor lock-in effects. But what is the technology behind it?

"Palantir is not a data company, it's a software company," the US technology company has been emphasizing for years. Palantir wants its software to be the central infrastructure for decision support for every organization. In order to achieve this, it is necessary to bring "order to the data chaos" for customers and create a networked organization. Palantir's business model is based on software-as-a-service, with service fees and training costs in addition to the license costs. Choosing Palantir means becoming dependent and moving your digital infrastructure to a proprietary ecosystem.

The use of Palantir means not only a technical, but also a structural commitment of your own organization to the US company. A subsequent system change is costly. Although Palantir offers numerous interfaces to integrate existing systems and data sources, there is no option to move the system to another provider once it has been set up. Certain data can be exported, but the actual capital – the semantic structures, logics, interactions – are deeply linked to Palantir's system and only work there. At the heart of the architecture is a uniform ontology, a formal model of meaning that defines how the organization's data is understood, linked and made operationally usable. Data-driven decisions and processes based on such a system can change the way an organization works.

"Missing Link"

What's missing: In the fast-paced world of technology, we often don't have time to sort through all the news and background information. At the weekend, we want to take this time to follow the side paths away from the current affairs, try out other perspectives and make nuances audible.

• All our the "Missing Links"

Palantir changes organizations

Whether Palantir is suitable as a nationwide interim solution for the police and should be introduced, as called for by the Federal Council in March 2025 and desired by some interior ministers, seems questionable. In the meantime, there is growing resistance to the US big data software for the police and no agreement could be reached at the Conference of Interior Ministers (IMK) for the nationwide introduction of Palantir. Instead, the IMK stated in its resolution of June 18 that "the ability of automated data analysis as a key element of the future digital security infrastructure must not be subject to any structural influence by non-European states with regard to availability, confidentiality, integrity and its legal conformity" and "against this background, the development of a digitally sovereign solution should be sought".

The federal states are still free to (further) cooperate with Palantir. Baden-Württemberg has been planning the introduction of a state VeRA for some time and, following consultation with Bavaria, has earmarked 18.5 million euros in budget funds for 2025/2026. Saxony-Anhalt also wants to introduce Palantir, reports MDR.

The experience of the Hesse state police, which has been using Palantir Gotham as "HessenData" since 2017, shows how the way of working with Palantir is changing. Every year, several thousand queries are carried out using the system, which can be accessed from any workstation. When the Federal Constitutional Court objected to the rules on data analysis in the police in February 2023, the state of Hesse was given a deadline to amend the law, during which the use of HessenData was still permitted under certain conditions. Hesse adapted its legal basis, which is why HessenData is still in operation. The system ensures greater efficiency and "every use in the day-to-day work of colleagues is actually a huge success", the SZ quotes a police officer from Innovation Hub 110, who is responsible for operating the platform, in a recent article. Palantir appears to be an integral part of Hesse's investigative work and the state is also interested in expanding its analysis capabilities with AI. In May 2025, employees from Innovation Hub 110 took part in a Palantir AIP Expo in Munich with their own live demo, as they announced on LinkedIn. Palantir organizes such AIP Expos as a closed networking event for executives from its customer base to have individual customers demonstrate the integration of AI into their products in practice.

In order to better understand the underlying system architecture, it is worth taking a look at the central Palantir products –, in particular the modular platforms Foundry and Gotham. They form the technical basis for applications such as HessenData , DAR (NRW) or VeRA (Bavaria), which could easily be supplemented with AI functionalities using AIP (Artificial Intelligence Platform).

Not a data platform, but an operating system

The main products are Palantir Foundry and Palantir Gotham, plus Palantir AIP and Palantir Apollo. All products are advertised as "platforms", whereby it is emphasized that they are not a data platform, but a modular operating platform and are intended to serve as an OS.

Foundry is advertised as an "ontology-based operating system for the modern enterprise", while the slogan "Your software is the weapon system" promotes Gotham as an "operating system for global decision-making" for authorities and organizations in the security sector.

The operating system is based on the "Palantir Data Store" as the physical/technical level of the storage structure for raw and transformation data and the ontology as the logical level of meaning for the data. The Palantir Data Store is part of Foundry. Data integration, modeling and analysis take place in Foundry.

From 100 fragmented systems to one Palantir system

The ontology is the central structural and organizational principle at the heart of the Palantir system, which becomes the semantic backbone of the organization using the system.

An ontology is a formal model that describes concepts, entities and their properties and relationships to each other in a standardized and structured way.

This creates a common world of meaning in a central system in which data flows together across systems from a wide variety of sources and formats and can be standardized, contextualized and analyzed.

The ontology creates a common language as an intermediate semantic layer across the heterogeneous IT system landscapes of an organization and its data. And this language can also be used in collaboration with other organizations if they use the same ontology.

Palantir's ontology creates a digital twin of an organization. Data, processes, rules and user interactions are brought together centrally and structured in a uniform model. This knowledge can be processed, searched and logically linked by machine. Not only can correlations be understood and conclusions drawn, but controllable actions can also be carried out.

This ontology is fully customizable and extensible – It forms the structure on which all analyses, data flows and user interfaces in Palantir products are based. The ontology determines what is visible, what is linked and what is considered relevant.

The core components of the Palantir ontology

• Object types – define entities such as "person", "thing", "event", "incident", "production line", "concept"
• Properties – describe properties/attributes of these entities, such as "name", "license plate", "location", "date", "timestamp"
• Shared properties – describe a property that is used on several object types, allowing property metadata to be managed centrally
• Link types – represent relationships between entities, for example "contains", "was with", "knows", "uses", "lives", "driver", "passenger", "is a member of", "witness", "accused", "suspect"
• Action types – allows changes to be made to the entities, their properties and relationships. Once the action has been carried out, further actions can also follow (automatically), which are stored in the respective action schemas.
• Functions – enable interaction with the entities in the form of logic that is executed
• Object views – Display of information and workflows relating to a specific object; for example, a profile could be displayed or a detailed display of all information associated with an object
• Roles – the direct application of roles to any ontology resource, regardless of the permissions on the original data source
• Interfaces – abstract schema for interfaces is an ontology type that describes the form of an object type and its functions. Interfaces enable consistent modeling and interaction with object types that have a common form.