With one week left until the tax filing deadline, experts are urging Americans to steer clear of consumer AI chatbots when preparing their returns.
Popular tools like ChatGPT, Gemini, Grok, and Claude may offer helpful explanations of tax concepts or brainstorming ideas, but tax professionals and federal authorities warn they pose significant risks for actual preparation and filing of taxes. Those risks include filing inaccurate returns, triggering audits, and facing penalties or additional taxes.
Privacy and Data Security Concerns
One major concern involves the handling of highly sensitive personal information. Inputting details such as Social Security numbers, W-2 forms, or bank records into general-purpose AI systems raises questions about data security and potential leaks. The National Institute of Standards and Technology (NIST) has highlighted how AI models can inadvertently leak, generate, or infer sensitive personal information by combining data from various sources, potentially exposing taxpayers’ details to the public domain.
High Error Rates in Tax Calculations
Consumer AI models also struggle with accuracy. They frequently misinterpret complex tax codes, misuse tax tables, and miscalculate eligibility for credits and deductions. A rigorous benchmark study known as TaxCalcBench , conducted by scholars and released through Column Tax, tested frontier models—including Gemini 2.5 Pro, Gemini 2.5 Flash, Claude Opus 4, and Claude Sonnet 4 on 51 realistic federal tax returns.
Even the best-performing model achieved only about 32% accuracy on fully correct returns, with consistent errors in areas like the Child Tax Credit and Earned Income Tax Credit, which involve intricate eligibility rules and phase-outs.
The IRS and its Taxpayer Advocate Service have explicitly cautioned against relying on AI-generated responses for complex tax questions or calculations, noting that the technology is still evolving and may not properly account for individual circumstances or the latest regulations.
The Problem of Hallucinations
AI “hallucinations”, where models fabricate plausible but incorrect information to fill gaps in their knowledge, also compound the issue. Studies on legal queries, which share similarities with tax law in their complexity and need for precision, have documented hallucination rates ranging from 58% to 82% for general-purpose chatbots, with even specialized legal AI tools erring 17% to 34% of the time.
Meanwhile, the IRS itself is deploying AI to scrutinize returns more effectively, increasing the likelihood that errors from consumer chatbots could draw unwanted attention.
In short, while AI chatbots can serve as useful educational aids for understanding tax rules, they are not a safe substitute for dedicated tax preparation software or qualified human professionals, at least not yet. Taxpayers concerned about accuracy and compliance are advised to use established tools from reputable providers or consult licensed tax preparers to minimize risks as the April 15th tax filing deadline approaches.
The post One Week Left: Ditch AI Chatbots For Filing Your Tax Returns appeared first on Loomered.
Comments are closed.