Exclusive: IRS Paid Palantir $1.8M to Build New Tool to Prioritize Tax Audits

Newly obtained public records acquired by WIRED through a public records request show the U.S. Internal Revenue Service (IRS) paid data analytics firm Palantir Technologies $1.8 million last year to upgrade a custom tool built to help the agency flag the highest-priority cases for audits, collection of unpaid taxes, and potential criminal investigations.

At the time the contract was signed, the IRS reported it relied on a patchwork of more than 100 separate business systems and 700 distinct workflows, built up over decades, to select cases where taxpayers may have misreported their tax obligations or owe outstanding funds to the government. As detecting potential tax mismatches has grown far more complex, the agency acknowledged its legacy infrastructure has become increasingly inefficient, leaving it in need of a targeted solution.

“This fragmented landscape can lead to a number of undesirable outcomes including but not limited to duplication of effort and cost, poor understanding of gaps in the coverage, and suboptimal case selection,” the IRS wrote in the contract scope document obtained by WIRED.

Palantir’s custom-built solution to this problem, named the Selection and Analytic Platform (SNAP), is designed to help the IRS streamline its process of identifying potential tax fraud cases. Per the documents, the software is still only being tested as part of an ongoing pilot program, and neither Palantir nor the IRS responded to WIRED’s requests for comment.

Government contracting records do not make clear how long Palantir has been developing SNAP, but the IRS has purchased Palantir-built technology since 2014. In total, Palantir has been awarded more than $200 million in contracts and committed payments from the IRS, and the documents show the agency is interested in deepening its working relationship with the firm.

It remains unclear exactly how SNAP will integrate with the IRS’s existing technology stack. Like most other Palantir tools, it is expected to layer on top of the IRS’s highly fragmented existing databases, helping human auditors spot red flags in tax filings that would otherwise go unnoticed. The contract confirms the IRS is turning to Palantir to support its broader push to modernize outdated internal software. Per one project document, the SNAP pilot is designed to pull “key information about contracts, vehicles and vendors” from unstructured data stored in supporting tax documents.

| Got a confidential tip? |

|---|

| Are you a current or former Palantir employee or government worker with insight into this project? We want to hear from you. Contact reporter Caroline Haskins securely via Signal at carolinehaskins.61, using a personal phone or computer unconnected to your employer. |

The IRS tasked Palantir with building three distinct case-selection models aligned with key sections of the current U.S. tax code. The three priority areas are: disaster zone claims (tax relief for natural disaster victims), Residential Clean Energy Credits (a program that offsets the cost of installing clean energy systems like solar panels or wind turbines), and Form 709 Gift Tax Returns (the form required when a taxpayer gifts high-value assets such as artwork, stocks, or private business holdings to another person).

Mitchell Gans, a Hofstra University professor specializing in gift and estate taxation, says that if SNAP is analyzing unstructured data from supporting documents, it is likely reviewing forms that include the legally required “adequate disclosure” of gifted property. IRS rules mandate these disclosures include a detailed breakdown of how the asset’s value was calculated, as well as information on the relationship between the gift giver and recipient.

For example, Gans explains, if a taxpayer gifts a private business to another person, their required disclosure must include supporting documentation for the business’s appraisal, such as “balance sheets and statements of net earnings, operating results, and dividends.”

Erica Neuman, an accounting and finance professor at Youngstown State University, adds that other types of public unstructured data could also be of interest to the IRS through SNAP, including transaction logs from peer-to-peer payment apps like Venmo, and public seller storefronts on e-commerce platforms like Etsy and Depop.

If SNAP does pull data from platforms like Venmo or Depop to select audit cases, that information would already have to be in the IRS’s possession. Contract language explicitly states the agency only wants Palantir to work with data that already exists in the SNAP system.

For decades, the IRS’s core method for prioritizing audit cases has relied on calculating a Discriminant Information Function (DIF) score for every taxpayer. According to the IRS, “the higher the score, the greater the audit potential.” Neuman notes the DIF scoring formula is a closely held black box, but researchers broadly agree the agency flags cases based on similarities between current filings and past returns that ultimately led to adjusted tax assessments after an audit.

Neuman has researched other alternative case-selection methods the IRS has tested in recent years, including contracting with platforms like Coinbase to analyze cryptocurrency transaction data, and scraping public social media posts for clues that an individual or business may be underreporting income.

The IRS has attempted to update its own in-house technology dozens of times, but Neuman says nearly all of these projects have been derailed by persistent technical problems. The agency “basically never had a full successful modernization since the 1960s,” she explains.

Other parts of the federal government have also struggled to modernize outdated infrastructure. For example, the process for current federal workers to apply for retirement still partially relies on paper documents stored in a limestone mine in rural Pennsylvania.

But Neuman says one unique barrier makes modernization especially difficult for the IRS: the agency is broadly unpopular with the U.S. public, making it a convenient punching bag for politicians looking to score political points with voters. As a result, there is often “a lack of political will to prioritize investing in the IRS,” she says.

During the Trump administration, this dynamic translated to shrinking budgets and widespread staffing cuts at the agency. As of February 2025, roughly 103,000 people worked for the IRS. By July 2025, more than 25,000 staff had resigned or accepted offers of deferred resignation and early retirement.

IRS leadership also tends to have extremely high turnover, which makes it hard for large, multiyear modernization projects to stay on track or reach completion. “The commissioner turnover is incredibly quick—and that, of course, has accelerated in the most recent years,” Neuman says. “But it's just hard to get things done when there's not enough resources, and no one is really pushing that hard for it.”