AI Professional Services: Company Liability for Generated Errors Rises

• Professionals and businesses are held responsible for errors generated by AI tools they use, not the AI developers.
• High-profile cases, including *Clinco v. Commissioner*, Deloitte's report, and Air Canada's chatbot, demonstrate the legal and financial risks of AI inaccuracies.
• An attorney in a U.S. Tax Court case faced potential sanctions for submitting a brief with three AI-generated, non-existent citations.
• Deloitte partially refunded fees after a 2025 report for the Australian government contained fabricated academic citations and a false quotation.
• Air Canada was ordered to pay 812 Canadian dollars in damages after its chatbot provided a customer with incorrect policy information in November 2022.

The use of AI tools does not transfer liability for errors to the AI developer; instead, the professional or company remains responsible for AI-generated inaccuracies.

The integration of artificial intelligence into daily professional workflows has seen a dramatic acceleration, transforming from a nascent technology into a ubiquitous tool. By February 2026, a significant 52% of all U.S. employees and an even higher 68% of professional services workers were leveraging AI, according to Gallup's findings. This trend is further underscored by Thomson Reuters' Spring 2026 report, indicating that 74% of professionals utilize AI multiple times weekly for work-related tasks. For fields such as accounting, law, and consulting, AI offers substantial advantages, including expediting research, synthesizing complex information, and aiding in communication and writing.

However, this rapid adoption carries a critical drawback: AI's capacity to generate information that, while appearing authoritative and polished, is fundamentally incorrect. When such AI-generated inaccuracies occur, the onus of responsibility for these errors falls squarely on the professional or business that relied on the AI's output, rather than on the developers of the AI systems themselves. This has already led to a range of negative repercussions, including lawsuits, court-imposed fines, significant financial losses, and public embarrassment stemming from fabricated citations, non-existent sources, and generally inaccurate AI-produced content.

Several high-profile incidents underscore the tangible risks associated with unverified AI outputs. In a notable case before the U.S. Tax Court, *Clinco v. Commissioner*, attorney Abraham Wagner submitted a brief containing citations that the court later determined were likely AI-generated "hallucinations," referring to non-existent Tax Court cases. The case involved a restaurant owner accused of underreporting over $2 million in business receipts and claiming unsupported depreciation deductions, with the core issue revolving around the taxpayer's accounting records. Although the court ultimately rejected the taxpayer's argument and warned Wagner about potential sanctions—including financial penalties or disciplinary action—it did not impose them in this instance. This case, examined in the September 2026 issue of the Journal of Accountancy, serves as a stark warning to legal professionals that AI hallucination legal risk will not be tolerated in court proceedings.

Beyond the courtroom, professional negligence generative AI has manifested in other sectors. In 2025, Deloitte, a global consulting firm, produced a 237-page report for the Australian government concerning an automated system for penalizing welfare recipients. This report was found to contain non-existent academic citations and a fabricated quotation attributed to a federal court judgment. Deloitte subsequently issued a revised version, attributing the errors to generative AI and agreeing to partially refund its fees, highlighting the direct financial consequences of Deloitte AI report errors. Similarly, the Air Canada chatbot liability case emerged in November 2022 when the airline's chatbot incorrectly informed a customer about a bereavement policy that allowed for post-purchase reimbursement. Air Canada refused to honor this non-existent policy, leading the customer to sue and successfully win 812 Canadian dollars in damages.

The recurring theme across these incidents is clear: the use of AI tools does not transfer liability for errors to the AI developer; instead, the professional or company remains responsible for AI-generated inaccuracies. This principle is paramount for lawyers and compliance officers, who must recognize that while AI can enhance efficiency, it does not absolve them of their professional duties. The potential for AI to produce "hallucinations"—convincing but false information—demands a rigorous approach to verification.

The implications for company liability AI professional services are profound. Professionals are expected to exercise independent judgment, and AI outputs must be treated as preliminary information requiring thorough human review and validation. Failing to implement stringent verification and oversight protocols for all AI outputs can lead to significant legal, financial, and reputational risks, particularly when advising clients or making critical business decisions. The Tax Court's warning in *Clinco v. Commissioner* underscores that courts will not excuse errors simply because they originated from an AI system, reinforcing the need for unwavering professional diligence.

Lawyers and compliance officers must recognize that the use of AI tools does not transfer liability for errors to the AI developer; instead, the professional or company remains responsible for AI-generated inaccuracies. This necessitates implementing stringent verification and oversight protocols for all AI outputs to mitigate legal, financial, and reputational risks, especially when advising clients or making critical business decisions.

Source: Original reporting via an industry publication.