We can’t keep regulating technology only after we have already implemented it and wait for courts to tell us years later where the limits should have been. 

Palantir’s software doesn’t need new data when the definition changes. It only needs a new question.

European countries have a decision to make. As governments search for alternatives to American technology, the debate around Palantir is becoming part of a much larger question about Europe’s technological sovereignty. In Germany, Hessen’s license for hessenDATA, the police analysis system based on Palantir’s Gotham software, is expected to expire in 2027. Bavaria, meanwhile, intends to continue using Palantir while it searches for an adequate alternative. But while much of the debate revolves around dependence on an American company and whether Europe can build something comparable, the debate risks confusing the provider with the power.

In around 2015, I first heard about Palantir while working a late-night bar shift in Frankfurt am Main and had an interesting conversation with two lawyers. They gloated about working on a framework that would allow Palantir’s software to be purchased and used in Hessen. After roughly explaining to me what the software actually is, it sounded quite dystopian and I doubted that it was lawful to use it. But their reply to me was actually depressingly simple: “Politicians have a blind spot for legality and are easy to lobby when the purpose is security. And there are always loopholes, it is literally our job to find them.”

Fast forward to Hessen in 2023. Palantir’s Gotham aka HessenDATA had been operating there since 2017, when the Federal Constitutional Court ruled that the relevant provision of Hessen’s police law was unconstitutional. The powers given to police to automatically analyse stored personal data were too broad and the requirements for using them too low. Seems like the so-called loophole, to implement it, use it and wait for the courts to catch up, had closed.

But there is a reason police want software like Palantir, and pretending otherwise makes the debate too easy. Governments already possess enormous amounts of information across different systems. The problem is making sense of it. Where you would need several police departments, research, late night work and an overdose on coffee to find patterns and therefore a conclusion, software can do parts of that work within a few clicks. 

Interestingly, even under the pretence of security, the public seem to at least not be terribly opposed to the software. A usual response is: “If you are not a criminal, you don’t have to worry about it.” But I don’t believe these people are naive. I think they came to their conclusion because the attention is focused on the motivation of the state and police, illegal data gathering and their own behaviour. If the purpose is to find criminals, the police acquired the information legally and I am not a criminal, why should I care?

The problem doesn’t end with how the data was gathered or why it is being used.

The scary bit about Palantir is in its analytical capabilities. It solves one huge problem for governments, not gathering data, but finding patterns within the data they already possess. Convenient, but it also outsources the process of turning information into meaning. A human brain is a pattern recognition machine itself, with the difference that once the context changes, our understanding of an identified pattern can change with it. Palantir organises those patterns through its Ontology, a model that defines objects, properties and relationships between them. Millions of disconnected pieces of information can suddenly become searchable relationships.

Los Angeles police used Palantir as part of Operation LASER, a program designed to identify people considered likely to be involved in future crime. Predictive Policing. But when the program was audited, almost half of those classified as “chronic offenders” had either never been arrested for a violent crime or had only one such arrest and 89% were people of color. The review found serious problems with how the program was implemented, and Operation LASER was discontinued in 2019. 

And this is where the Constitutional Court ruling in Hessen matters. The problem is not simply what data Hessen’s police possesses, but what automated analysis allows them to create from it. The Court warned that combining existing information could produce new knowledge, including behavioral, relationship and movement profiles, even reaching into entire groups and social environments.

And back to the argument that if you are not a criminal, you have nothing to worry about. What exactly is a criminal? Who defines that? We treat criminality, threat or extremism as permanent categories, laws change, governments change and political priorities change. What was protected political expression yesterday can become grounds for state scrutiny tomorrow. 

Palantir doesn’t need new data when the definition changes. It only needs a new question.

This is what makes the current debate in Europe so misleading. If the replacement has the same analytical capabilities, the same constitutional and democratic questions remain. We can’t keep regulating technology only after we have already implemented it and wait for courts to tell us years later where the limits should have been. 

There simply shouldn’t be any loophole for misuse to begin with.

Join the Palantir Pledge!

Progressive International has launched the Palantir Pledge: a commitment by public officials at all levels — city councilors and mayors, parliamentarians and ministers — to purge Palantir from the state, adopt ethical policies for procurement and investment, and protect our fundamental rights.

To find out more and sign the pledge, click here.

Do you want to be informed of DiEM25's actions? Sign up here