The IRS Wants Smarter Audits. Palantir Could Help Decide Who Gets Flagged.
The Internal Revenue Service (IRS) is embarking on a significant modernization effort, turning to the prominent data analytics firm Palantir for assistance in revolutionizing its audit selection process. Last year, the IRS allocated $1.8 million to Palantir, a sum directed towards enhancing a specialized tool designed to pinpoint "highest-value" cases for audits, collection of unpaid taxes, and potential criminal investigations. This development, revealed through public records obtained by WIRED, underscores the tax agency’s urgent need to upgrade its archaic systems and adopt a more data-driven approach to enforcement.
For decades, the IRS has grappled with an intricate and increasingly inefficient operational landscape. The agency disclosed that it relied on "more than 100 business systems and 700 methods," a sprawling infrastructure painstakingly built over many years, to select cases involving potential tax discrepancies or outstanding debts. As the complexity of financial transactions and tax reporting continued to grow, these legacy systems proved to be an overwhelming impediment. The IRS candidly admitted that this "fragmented landscape" led to a multitude of undesirable outcomes, including "duplication of effort and cost, poor understanding of gaps in the coverage, and suboptimal case selection." Such inefficiencies not only drained resources but also hindered the agency’s ability to effectively pursue tax evasion and ensure compliance, ultimately impacting government revenue.
To confront these deep-seated challenges, Palantir was contracted to develop a custom solution known as the "Selection and Analytic Platform," or SNAP. This innovative software is engineered to streamline the IRS’s process for identifying potential fraud cases, moving beyond the cumbersome manual methods and disparate databases that have long plagued the agency. Currently, SNAP is being deployed as part of a pilot program, indicating a cautious yet determined step towards integrating advanced analytics into the core functions of tax enforcement. Neither Palantir nor the IRS has publicly commented on the details of the pilot or the broader implications of this collaboration.
While the precise duration of Palantir’s involvement with SNAP remains undisclosed, government contracting records reveal a long-standing relationship between the two entities, dating back to 2014. Over this period, Palantir has secured contracts and obligated payments from the IRS totaling more than $200 million. The recent investment in SNAP, coupled with the IRS’s stated interest in deepening its engagement with Palantir, signals a clear strategic direction: a commitment to leveraging sophisticated technology to modernize its operations and enhance its capacity for targeted enforcement.
The integration of SNAP into the IRS’s existing technological ecosystem is a complex undertaking. Given the highly splintered nature of the agency’s databases, SNAP, much like other Palantir tools, is likely designed to act as an overlay, consolidating and analyzing data from disparate sources. Its primary function is to assist human auditors by identifying "red flags" in tax filings that might otherwise go unnoticed. By sifting through vast quantities of information, the platform aims to highlight anomalies, patterns, or discrepancies that warrant further investigation, thereby empowering auditors to focus their efforts on the most promising cases. One document obtained by WIRED specifies that Palantir’s SNAP pilot is specifically tasked with surfacing "key information about contracts, vehicles and vendors" from "unstructured data from supporting documents," suggesting its capability to process and derive insights from a wide range of textual and non-numeric data.
The pilot program focuses on three distinct "case selection methods" directly related to specific aspects of the existing tax code, illustrating the practical applications of SNAP. These areas were chosen to test the platform’s efficacy in diverse and potentially complex scenarios:
- Disaster Zone Claims: This category involves tax relief provisions for victims of natural disasters. While intended to provide crucial support, such claims can sometimes be susceptible to fraud, with individuals potentially exaggerating losses or fabricating claims. SNAP could analyze patterns in claims, cross-reference them with disaster declarations, and flag unusual filings.
- Residential Clean Energy Credits: This program offers tax credits to incentivize homeowners to install renewable energy systems like solar panels or wind turbines. The complexity of valuations, installation costs, and eligibility criteria can create opportunities for inflated claims or fraudulent applications. SNAP could scrutinize supporting documentation, compare costs against market rates, and identify suspicious credit requests.
- Form 709 Gift Tax Returns: This form must be filed when individuals give away valuable assets, such as artwork, stocks, or interests in corporate entities, exceeding certain thresholds. The valuation of non-liquid or complex assets is often subjective and can be a common area for underreporting or misrepresentation.
Mitchell Gans, a distinguished professor at Hofstra University specializing in gift and estate taxes, offers valuable insight into how SNAP might approach Form 709. He suggests that if SNAP is indeed analyzing unstructured data from supporting documents, it would likely be examining forms providing "adequate disclosure" of gifted property. The IRS mandates that such disclosures include "a detailed description" of how the property’s value was determined, along with the relationship between the giver and the recipient. For instance, if a private business is gifted, the disclosure would necessitate extensive supporting information, such as "balance sheets and statements of net earnings, operating results, and dividends," to substantiate the valuation. SNAP’s ability to process and interpret these complex financial documents could significantly enhance the IRS’s capacity to identify undervalued gifts.
Erica Neuman, an accounting and finance professor at Youngstown State University, further expands on the concept of unstructured data, pointing to public logs from money transfer applications like Venmo, as well as public storefronts on e-commerce platforms such as Etsy and Depop, as potential sources of relevant information for the IRS. These platforms contain vast amounts of transactional data, user profiles, and product descriptions that, if accessible and analyzed, could reveal undeclared income or suspicious financial activities. However, it is crucial to note a significant constraint outlined in the contract documents: Palantir is explicitly instructed to use only "existing data in SNAP today." This means that while data from Venmo or Depop might theoretically be valuable, SNAP’s current scope is limited to data already possessed and integrated by the IRS. The platform is designed to make sense of the agency’s internal, fragmented data rather than actively acquiring new external data sources, at least for this pilot phase.
The IRS’s move to embrace Palantir’s advanced analytics signals a broader trend in government agencies utilizing big data and artificial intelligence to enhance operational efficiency and enforcement capabilities. However, this advancement is not without its complexities and potential concerns. The use of AI in auditing raises critical questions about fairness, bias, and privacy. While proponents argue that data-driven audits can be more objective and efficient, critics often express apprehension that algorithmic systems might inadvertently perpetuate or amplify existing biases, potentially targeting specific demographics or income groups disproportionately. Ensuring transparency in how these algorithms operate and establishing robust oversight mechanisms will be paramount to maintaining public trust.
Furthermore, given Palantir’s history with sensitive government contracts, often involving intelligence and defense agencies, its increasing role in domestic tax enforcement naturally sparks privacy debates. While the current contract emphasizes the use of "existing data," the public perception of a powerful data analytics firm sifting through personal financial information underscores the need for stringent data governance and clear ethical guidelines.
Ultimately, the pilot program for SNAP represents a pivotal moment for the IRS. If successful, it could usher in an era of smarter, more targeted tax enforcement, allowing the agency to recover billions in unpaid taxes and crack down on criminal activities more effectively. It signifies a long-overdue modernization of an agency critical to the nation’s financial health. The partnership with Palantir underscores the growing reliance of governmental bodies on cutting-edge technology to navigate the complexities of the digital age, while also highlighting the ongoing challenge of balancing technological advancement with principles of fairness, transparency, and privacy in public service. The outcome of this pilot will likely shape the future trajectory of tax audits and the broader role of AI in government operations for years to come.
