The IRS announcement that it will soon be deploying artificial intelligence (AI) software to enforce the U.S. tax code has raised concerns among cyber experts that without proper oversight, the new technology could be weaponized against conservative groups.
In remarks made April 17 at the UiPath on Tour: Public Sector event in Washington, IRS Commissioner Danny Werfel announced the agency would be utilizing the tools to better assist IRS agents in identifying potential tax cheats.
“Right now, I believe that there are AI solutions that we have not yet leveraged that exist today that can help with some of these basic questions to the benefit of taxpayers,” Mr. Werfel said. “And on the other side of the equation, we are using AI today to do even more to unlock and spot this complexity.”
The government’s implementation of the growing technology will be instrumental in helping the agency uncover the most elaborate tax schemes, according to Mr. Werfel.
“I also think of a chess analogy to where we are going to be assessing some of the complexity of how money is moved across different subsidiaries, into tax shelters and holding companies,” Mr. Werfel said. “Sometimes [it’s] done completely legally, and sometimes, unfortunately, it’s done illegally.”
“A human makes the decision, ultimately, when we do these types of efforts on the enforcement side, but we want the computer essentially helping us be a better chess player,” he added.
However, tech expert Jake Denton, a research associate at The Heritage Foundation’s Tech Policy Center, told The Epoch Times that taxpayers should be skeptical of how the agency will use the technology.
“This technology really gives the IRS a shield to bulk flag anyone it finds problematic, which is very troubling,” said Mr. Denton. “It is encouraged under this model to use broad strokes to identify individuals and can then, if asked, lean into the supposedly all knowing AI model under the claim that it is proprietary, and therefore, escape any potential accountability.
Previous Discrimination Against Conservatives
In 2017, the IRS was forced to express “its sincere apology” after it was discovered that the agency had targeted conservative organizations, including Linchpins of Liberty—along with 40 other conservative groups—in their applications for tax-exempt status. Later that year the Department of Justice would go on to settle a class action suit that included 428 members who alleged they were victims of political discrimination by authorities within the IRS.In a statement announcing the settlement, former Attorney General Jeff Sessions said that the agency had betrayed the public trust.
“Chief Justice John Marshall wrote ’that the power to tax involves the power to destroy … [is] not to be denied.‘ And it should also be without question that our First Amendment prohibits the federal government from treating groups differently based solely on their viewpoint or ideology,” said Mr. Sessions. “But it is now clear that during the last Administration, the IRS began using inappropriate criteria to screen applications for 501(c) status. These criteria included names such as ’Tea Party,‘ ’Patriots,‘ or ’9/12‘ or policy positions concerning government spending or taxes, education of the public to ’make America a better place to live,’ or statements criticizing how the country was being run. It is also clear these criteria disproportionately impacted conservative groups.
“As a result of these criteria, the IRS transferred hundreds of applications to a specifically designated group of IRS agents for additional levels of review, questioning and delay. In many instances, the IRS then requested highly sensitive information from applicants, such as donor information, that was not needed to make a determination of tax-exempt status.
AI Must Be Open Source
Despite the abuse of power, in 2022, the Democrats’ Inflation Reduction Act rewarded the IRS with a large cash infusion, delivering nearly $80 billion, a portion of which has been earmarked for the creation and implementation of the new technology.Mr. Denton warns that the new AI technology gives the agency a significantly more powerful tool to covertly discriminate against certain groups than anything that could have been imagined in 2017.
“This kind of AI integration is perfect to cultivate the kind of discriminatory corruption people are most concerned about,” said Mr. Denton. “AI would allow bad actors within the agency to use the technology to target certain groups and now there would be nothing that the individual could do to contest the models since the process would have been automated away to an unexplainable black box.”
“In the old model you could at least bring forward the person who flagged you and at least have the right to challenge your accuser,” he added.
Any hopes of reigning in the powerful technology and averting potential abuse will be contingent on the public’s demand for transparency, according to Mr. Denton.
“We don’t want to block the IRS from acquiring AI,” said Mr. Denton. “The best thing would be to optimize the IRS completely away. But the answer requires a certain amount of transparency. If they are going to bring in the model it has to be open sourced, and explainable.”
“We can set these kinds of guidelines before it is fully adopted and given its history, shouldn’t accept anything less,” he added.
