Artificial intelligence is transforming the typical workday in a variety of industries. The Internal Revenue Service (IRS) is no different. For years, the agency has been working with Palantir Technologies to better process tax returns and identify potential audit candidates.
While the IRS is still a relatively large agency, the number of tax returns it must process every year is staggering. In addition, the IRS has significantly fewer employees than a decade ago, particularly in enforcement roles like revenue agents. The IRS has never been able to audit all tax returns with potential issues, and its workforce reduction has led to audits at near-historic lows. Unsurprisingly, then, the IRS is increasingly focusing on audits that have the “highest value” (i.e., audits that will result in the IRS getting the most money).
That’s where AI and Palantir come in. Co-founded by Alex Karp and Peter Thiel, Palantir provides numerous data-integration and analytics platforms to government agencies worldwide. Federal agencies are Palantir's largest customer group, with government contracts accounting for approximately half of its revenue. It has long provided the government with data platforms for counterterrorism and border security operations in the U.S. Earlier this year, the Department of Homeland Security awarded Palantir a five-year contract worth up to $1 billion.
Now, according to reporting in Wired, the IRS is exploring an increasing role for the data analytics giant in processing tax returns, conducting criminal investigations, and conducting audits.
How the IRS Is Using Palantir
The Selection and Analytic Platform (SNAP) is an AI tool meant to identify the highest-value cases for audits. It is currently part of a pilot program designed to streamline information from over 100 separate business systems. The idea is to let IRS analysts see relationships across taxpayers, entities, assets, and transactions.
This is not the first time the agency has partnered with Palantir. All told, Palantir has been awarded IRS contracts worth approximately $200 million since 2014. The IRS awarded up to $11.8 million in January 2025 for the SNAP program and other Palantir tools, and recently allocated just under 10% of that budget to enhance SNAP’s capabilities.
Does the IRS’s Adoption of AI Mean Audits Are More Likely?
Not necessarily, and not this year in particular. It is still in its pilot program phase. Even if SNAP use expands, it’s likely the main focus will remain on large-scale fraud, not on expanding random audits of taxpayers. The IRS intends to use Palantir’s platforms to identify patterns and connections that would be hard to spot manually, like transaction flows that suggest tax evasion schemes.
This could mean analyzing transactions on payment apps like Venmo, among numerous other potential transactions. The IRS would already need to possess all the information SNAP uses. The contract specifies that SNAP can only analyze existing data.
Historically, the IRS has used the Discriminant Information Function (DIF) score to determine whether a tax return should be audited. No one outside of the IRS knows how the DIF score is calculated (to avoid helping taxpayers game the system), but it appears the IRS will use SNAP alongside existing systems like DIF, not replace it.
What About Those Privacy Concerns?
The benefits the IRS sees in streamlining the process to determine which returns to audit seem straightforward. However, some privacy advocates are warning about the risk of increased financial surveillance. Concerns are always raised when the federal government is using algorithms to compile data on Americans. So, is this simply a modernization effort by the IRS to get rid of obsolete legacy systems? Or is it a red flag suggesting an increasing creep toward a surveillance state?
As with the DIF, the IRS’s use of SNAP is essentially a black box. It’s not clear if the IRS is on a path to create a de facto searchable “mega database” that could be used to compile extensive federal records (or even quasi-public transaction data) on American taxpayers. Even if not, there is concern that SNAP may violate the Privacy Act of 1974.
On the other hand, previous modernization efforts at the IRS have had mixed success, largely because the IRS, despite being a nonpartisan institution, often has its priorities shifted by the current administration. The Trump administration has embraced the use of AI and other technologies across all aspects of government, but future administrations may be more leery of working with tech leaders such as Peter Thiel and Elon Musk. Musk headed the Department of Government Efficiency, which kept a hiring freeze at the IRS and had 17,500 workers leave under deferred resignation programs.
For now, it remains to be seen how much the IRS will integrate Palantir tools into its existing processes.
Related Resources
- Federal Judge’s Injunction Halts IRS’s Taxpayer Data-Sharing With DHS and ICE (FindLaw’s Law and Daily Life)
- IRS Audit Representation (Findlaw’s Learn About the Law)
- New for the 2026 Tax Season: The Car Loan Interest Deduction (FindLaw’s Law and Daily Life)