Palantir and Credal AI: US Health Department Screens Grants for 'Correctness'

Imagine applying for a research grant after spending months on studies and calculations. Your application is rejected in a fraction of a second, not because the math is bad, but because an algorithm found "wrong" words in the text. That is exactly what is happening now in the corridors of the US Department of Health and Human Services (HHS). Since March 2025, the agency has officially put neural networks on guard over ideological purity, using Palantir's computing power and tools from the startup Credal AI.

This is not just another story about software rollout. It is a tectonic shift in how the state interacts with science and society. Checking for compliance with political directives used to be slow and human. Now Palantir, a company whose image is inseparable from Peter Thiel and government surveillance, has supplied the infrastructure for automated censorship. Working alongside it is Credal AI, a startup that specializes in safely deploying language models in corporate processes. Together they have built an ideal filter for scrubbing a particular agenda from the public budget.

Two acronyms and one broad concept are in the crosshairs: DEI (diversity, equity and inclusion) and so-called "gender ideology." If your health program accounts for the specific needs of minorities or uses terminology the agency's current leadership considers excessive, your chances of getting funding approach zero. The algorithm does not grasp context. It looks for markers flagged as "undesirable" in the current political cycle.

To grasp the scale, recall what Palantir did before. Its Foundry platform has for years helped intelligence services link scattered data into a single picture to find threats or analyze supply chains. Repurposing that power to hunt for "social engineering" in grants looks like a logical, if frightening, evolution of control technology. AI used to be accused of a liberal bias; now the pendulum has swung hard the other way, turning it into a tool of conservative revision.

The problem is not only political but also rooted in the nature of large language models. We know neural networks can make mistakes and misread text. When such a model becomes a judge deciding the fate of millions of dollars, transparency in the process vanishes. How exactly did Credal AI configure its filters? What weights are assigned to the words "equality" or "gender"? The answers are hidden behind corporate secrecy and departmental classification markings.

For the scientific community, this means the start of an era of self-censorship. Researchers whose work depends on grants for survival will now think three times before using certain vocabulary. This sets a dangerous precedent: science begins to adapt to an algorithmic filter tuned to the current political climate. Today it is the fight against DEI, but what will the neural networks be told to look for tomorrow?

This case shows plainly that the era of "neutral AI" ended before it ever began. Technology has become an extension of the will of those who hold the budget lever. The use of Palantir in this scheme only confirms that effective data management in the modern world is, above all, effective filtering of unwelcome meanings.

Bottom line: Will the "ideological AI filter" become the standard for all government bodies, or will we see large-scale legal resistance from the scientific community?