Who's Liable When AI Makes Mistakes? Companies Face Growing Accountability Crisis

AI Adoption Accelerates Faster Than Accountability Frameworks

Artificial intelligence has rapidly transitioned from experimental technology to essential workplace infrastructure. By February 2026, 52% of U.S. employees and 68% of professional services workers were using AI, according to Gallup1

1

. Thomson Reuters reported that 74% of professionals use AI several times weekly for work-related tasks1

1

. For accountants, attorneys, consultants and other professionals, AI accelerates research, summarizes complex information, and assists with communication. Yet this widespread adoption has created an accountability crisis in businesses as AI mistakes increasingly lead to legal liability, financial consequences, and reputational damage.

Courts Hold Companies Responsible for AI Errors

When AI makes a mistake, companies may be held liable rather than AI developers. The U.S. Tax Court case Clinco v. Commissioner demonstrates how AI harms clients and customers through professional negligence. Attorney Abraham Wagner submitted a brief containing citations to four cases, three of which the court determined were AI hallucinations citing nonexistent Tax Court cases1

1

. The court warned that submitting fake citations could lead to sanctions including financial penalties or disciplinary action, establishing that AI-generated misinformation will not be tolerated in legal proceedings.

Air Canada learned this lesson expensively when its chatbot told a customer in November 2022 that he could purchase a ticket under a bereavement policy and receive reimbursement later—contradicting actual company policy. The British Columbia Civil Resolution Tribunal awarded the customer 812 Canadian dollars in damages, ruling that Air Canada had failed to take reasonable care ensuring its AI bot provided accurate information and had negligently misrepresented the bereavement policy1

1

. This Air Canada chatbot case established clear professional liability for AI-driven operational risks.

Global Consulting Firms Face AI Accountability Issues

Deloitte's experience illustrates how AI errors extend beyond small businesses to major consulting firms. When Deloitte produced a 237-page report in 2025 for the Australian government examining automated welfare penalty systems, the document contained nonexistent academic citations and a fabricated quotation attributed to a federal court judgment1

1

. The global consulting firm issued a revised version, blamed generative AI for the errors, and agreed to partially refund its fees—demonstrating that even established professional services firms cannot escape legal liability when AI hallucinations and errors compromise client deliverables.

Managers Cannot Agree on Who Is at Fault When AI Makes a Mistake

A Wharton School study surveying over 1,000 business leaders reveals profound confusion about AI accountability. When asked who should answer for consequential AI mistakes, approximately one-quarter pointed to the employee using the tool, another quarter named that employee's manager, and a similar share chose the company's risk team2

2

. Around 10% blamed the AI vendor, while about the same percentage were unsure2

2

. This fragmented view of responsibility creates managerial challenges as businesses struggle to establish clear accountability frameworks.

Real-World Examples Show Escalating Operational Risks

Leeron Walters, a vice president at employee-monitoring software company Teramind, documented cases where simple prompt mistakes created massive problems. At a major U.S. retailer, an employee used AI to automate customer emails, but a mistake in the bot's instructions caused it to send false refund and cancellation messages to more than 18,000 customers in under six hours, generating thousands of service requests2

2

. At a U.S. consumer electronics company, a marketing employee used AI to set product prices, but a prompt error listed a flagship device at an 85% discount for three hours. More than 700 customers placed orders before the company corrected the mistake, leaving the customer-service team managing weeks of fallout2

2

.

New AI Agents Increase Stakes for Companies

Companies like Google, OpenAI and Meta are releasing AI agents with expanded capabilities that amplify potential consequences of AI mistakes. OpenAI introduced Dots late last month—always-on AI agents that can connect to over 4,000 apps and continue working on projects between user check-ins2

2

. OpenAI is also developing workplace versions with their own credentials and access to company systems. Google announced its Gemini agent for enterprise users, capable of planning and completing multistep work across company systems including drafting content and writing code2

2

. Meta's Muse personal agent manages email and travel booking tasks while continuing work after users leave the app, reaching more than 2.5 million downloads in its first two weeks according to Sensor Tower data2

2

.

Why AI Hallucinations Are Difficult to Detect

AI errors present unique detection challenges compared to human mistakes. Unlike obvious human errors, AI-generated misinformation may include supporting explanations, citations, or references that make passages appear complete and credible1

1

. This polished presentation of incorrect information creates particular risks for professionals whose work requires authoritative judgment. The National Institute of Standards and Technology issued a report on AI risk management addressing these concerns, though the technology's rapid advancement continues to outpace regulatory frameworks.

What Businesses Should Watch

As AI liability cases accumulate, businesses face mounting pressure to establish clear accountability structures before courts impose them. IBM research shows 41% of consumers use AI assistants to research products while 31% use them to find deals2

2

, indicating AI's expanding role in customer-facing operations. Organizations must determine who bears responsibility when AI makes a mistake, implement verification processes for AI outputs, and recognize that legal liability ultimately rests with the company—not the AI vendor. The question is no longer whether AI accountability matters, but how quickly businesses can develop frameworks that protect clients, customers, and their own operations from AI-driven operational risks.