Experts warn against using AI to prepare your taxes

An accounting professor warns that AI chatbots should not be used to prepare tax returns because they may fabricate information and lack privacy protections. The key risk is legal exposure: taxpayers remain responsible for any errors on filed returns. The article is advisory in nature and is unlikely to have a meaningful market impact.

Analysis

This is less about consumer tax prep and more about a slow-moving liability reset for any software that embeds generative AI into regulated workflows. The market is still pricing AI as a productivity layer, but in tax, legal, and compliance use cases the marginal output is often not just noisy — it can create non-delegable liability for the user while leaving the vendor with reputational and, eventually, product-design risk. That asymmetry favors firms selling deterministic workflow software, audit trails, and human-review augmentation over pure chatbot UX.

The second-order winner set is cybersecurity/privacy tooling and enterprise governance software. If users become more aware that prompts may contain SSNs, wage data, and banking details, adoption of consumer-grade AI for sensitive tasks should slow over the next 1-2 filing seasons, while demand for data-loss prevention, identity protection, and enterprise prompt logging rises. That is a subtle but important shift: the negative sentiment on consumer AI assistants can coexist with incremental revenue upside for vendors that help companies prove compliance and contain model usage.

From a trade perspective, the near-term catalyst is behavioral rather than regulatory: one high-profile tax error or privacy incident would likely cause a sharp but temporary pullback in consumer AI usage and a rotation into “safe AI” names. The longer-dated risk is that regulators start treating AI tax prep like other advice categories, which would force human-in-the-loop disclosure and documentation requirements over 6-18 months, raising the cost of distribution for consumer-facing AI apps. The contrarian view is that the headline risk may be overdone for large incumbents, because most of their enterprise deals already include indemnities, logging, and closed-data environments; the real fragility sits with standalone AI-native point solutions that rely on trust rather than process control.

Key Decisions for Investors

• Short consumer-facing AI assistants that market themselves as general-purpose advisors; use a 1-3 month horizon and size for headline volatility rather than a secular thesis.
• Go long cybersecurity/privacy enablers such as CRWD, PANW, and ZS over the next 2-6 months; the risk/reward improves if regulatory concern drives enterprise spending on data controls and prompt governance.
• Pair trade: long enterprise workflow software with embedded AI controls (e.g., INTU / ADP) versus short unprofitable AI-native consumer app exposure, targeting a 6-12 month outperformance as compliance wins over novelty.
• Buy calls on identity protection/data privacy names into tax season and earnings season, when consumer anxiety around sensitive data typically peaks; use defined-risk upside exposure rather than common stock.
• Avoid chasing pure-play AI enthusiasm on consumer use-case announcements until there is evidence of human-review integration and auditability; the first legal incident can create a 20-30% drawdown in vulnerable names.