The potential threat of “especially deceptive” tax scams created using artificial intelligence chatbots like ChatGPT has caught the interest of a bipartisan group of Senate taxwriters.
Four Senate Finance Committee members, led by Sen. Maggie Hassan, D-N.H., called on the IRS to “use all the tools at its disposal” to counter AI-generated tax scams in a May 1 letter to the agency.
The letter said that an AI program like ChatGPT could trick taxpayers into paying fake fines and taxes more easily because of its ability to avoid obvious red flags found in other scam messages, such as spelling mistakes and grammatical errors.
“By contrast, tax scams generated by new AI tools are professionally composed and specifically tailored to trick vulnerable taxpayers,” according to the letter, which was also signed by Senate Finance Committee Chair Ron Wyden, D-Ore., and members James Lankford, R-Okla., and Chuck Grassley, R-Iowa.
The lawmakers requested answers on how the IRS plans to address a possible increase in scams using AI tools, how it plans to educate taxpayers and tax professionals on the matter, and whether the IRS has received any reports of AI-generated scams.
Hassan had asked IRS Commissioner Daniel Werfel about how the agency would handle the dangers of ChatGPT during an April 19 hearing, saying, “These AI tools could be used to quickly generate many well-composed, convincing messages that trick taxpayers into providing scammers with personal financial information.”
Werfel said the IRS hopes to use the additional funds it receives from the Inflation Reduction Act (P.L. 117-169) to ensure AI and other emerging technologies don’t outpace the agency as it works to protect taxpayers from scams.
Werfel said that he found the IRS “to be in very good shape in terms of our understanding of the portfolio of fraud that’s out there and how we detect it. But we have more work to do.”
Researchers with the security firm WithSecure found in a January study that ChatGPT, which was publicly released in November 2022, gives cybercriminals “the ability to make any targeted communication as part of an attack more effective.” Security firm Check Point Research also tested at the end of 2022 how the large language model can create phishing emails.
The chatbot could strengthen existing tax scams conducted over email and on social media, many of which have been highlighted in the IRS’s “Dirty Dozen” list of warnings.
The potential for AI as a tool in the tax industry isn’t all malicious. The IRS has explored its use to help with compliance and taxpayer service and training, and programs like ChatGPT could automate legal advice and tax return preparation.