AI Errors in Deloitte Report Trigger Refund, Expose Growing Risk

Deloitte will partially refund the Australian government for a report after officials identified apparent AI-generated errors in the consulting firm’s document, according to reports. The 237-page, $290,000 report for the country’s Department of Employment and Workplace Relations included a fabricated quote attributed to a federal court judgment and cited academic research papers that did not exist, Fortune reported in October. 

The incident involving the Big Four accounting and consulting firm has drawn international attention, with some experts saying the situation should be a “wake-up call” for companies using artificial intelligence platforms without guardrails in place, CFO Dive reported. 

The increasing use of generative-technology tools by professionals is a serious risk across nearly all industries, said Josh Zack, Broker, Professional Liability, Burns & Wilcox, Chicago, Illinois. 

“It is not just this one occurrence. We are seeing errors come up more frequently with the emergence of AI in professional services and not just in tech firms,” Zack said. “A major firm like Deloitte using AI and generating these AI hallucinations and information that does not exist presents a clear and present exposure.” 

It is not just this one occurrence. We are seeing errors come up more frequently with the emergence of AI in professional services and not just in tech firms.

Errors in professional services caused by the misuse of AI could prompt a variety of lawsuits, potentially triggering coverage under Errors & Omissions (E&O) Insurance or Directors & Officers (D&O) Insurance, Zack said. Both the vendor providing the professional services, and the organizations receiving those services, should carry these policies. 

“I think there are some D&O and E&O exposures on both sides,” said Ryan Ascenzo, Senior Broker, Professional Liability, Burns & Wilcox, Brokerage Division, New York, New York. “If [the Deloitte case] happened here in the U.S., and the accounting firm did this for a publicly traded company, there could be shareholder actions against the board of directors asking about what kind of due diligence they are doing on the vendors that provide these reports. Then for the service provider, there is an E&O issue, as they performed a service and did not meet expectations.” 

The escalating risk of AI-generated errors 

According to data from a new McKinsey survey on the state of AI, nearly 9 out of 10 organizations report using AI regularly, with about two-thirds indicating that they are in the experimentation or piloting phase. About 88% of organizations reported using AI in at least one business function, up from 78% a year ago, McKinsey & Company noted in its Nov. 5 report. 

Despite ongoing concerns about AI’s accuracy — for example, recent data from the Pew Research Center shows that 66% of adults overall and 70% of experts are highly concerned about inaccurate information from AI — many companies have not yet established clear safeguards around the use of AI. According to Forbes, 65% of companies did not have policies in place regarding the use of generative AI by partners and suppliers as of November 2024. 

“We are putting so much trust in AI, and who knows whether that data is accurate? In a lot of cases, the right checks and balances are not in place to make sure the data is accurate,” Zack said. 

We are putting so much trust in AI, and who knows whether that data is accurate? In a lot of cases, the right checks and balances are not in place to make sure the data is accurate.

As AI becomes more embedded in professional workflows, inaccuracies generated by automated tools could introduce significant legal and financial risks. The extent and cost of the damage these errors could cause remains to be seen, Ascenzo said. 

“I do think AI is creating a lot of errors and omissions as we speak. It just has not necessarily been uncovered yet or has not been at a large scale,” he said. “Just like the old saying with [cyberattacks], it is not a matter of if, but when.” 

I do think AI is creating a lot of errors and omissions as we speak, it just has not necessarily been uncovered yet or has not been at a large scale.

How insurance policies may respond  

E&O Insurance can help organizations respond to allegations that their professional services contained errors or omissions that caused financial loss, with the policy potentially covering legal defense, settlements, and other expenses. Claims under E&O Insurance policies often hinge on the presence of financial loss, Ascenzo noted. While fees that need to be refunded are generally not insurable, “damages beyond the fees can be,” he said. 

However, organizations should find out how their policy would respond to AI-related inaccuracies, Zack said. “More carriers today are including affirmative AI coverage on their policies, so if AI is misused, there is coverage,” he explained. On the other hand, certain industries, such as architects and engineers, may have exclusions limiting protection for AI tools, he said. “It is important to look out for those endorsements and any fine print excluding AI.” 

The same advice applies to coverage under D&O Insurance, which is designed to protect a company’s leaders against claims of mismanagement or oversight failures that result in financial harm. Many D&O Insurance policies have a cyber exclusion, Ascenzo said, which may or may not apply to AI risks. “There is a lot of differentiation from carriers on that cyber exclusion,” he said, adding that possible D&O Insurance exclusions on professional services and intellectual property should also be reviewed. 

“Business decisions that do not go according to plan often turn into D&O claims,” Ascenzo added, pointing to concerns shareholders may have over AI-related decisions following news of an error. “If stock drops because of the news, it could bubble up to be something much larger, such as a securities class-action suit.” 

Depending on the allegations, other policies could also be important. Media Liability Insurance could respond to situations where AI is “used to create images that could infringe on copyright,” Zack said, while Cyber & Privacy Liability Insurance may address claims involving data breaches linked to AI use. 

Experts advise ‘transparency and honesty’ regarding AI 

As companies integrate AI tools into research and other day-to-day operations, experts emphasize that even well-run organizations can face unexpected losses. Some business owners assume they do not need insurance because they have “never had an issue,” Ascenzo said. “That is all self-perception,” he said. “All it takes is someone outside your sphere to create some turmoil or cloudiness and all of a sudden you may be facing a claim.” 

Clear communication about how AI is used is becoming increasingly important, he added. “I think transparency and honesty is still the best policy upfront,” Ascenzo said. “Speak with your legal team about disclaimers, ask them how it should be disclosed to stakeholders and clients, and reach out to service partners that do AI fact-checking. This is an emerging exposure, and I do not think anyone has a silver bullet on how to eliminate the risk altogether.” 

I think transparency and honesty is still the best policy upfront. Speak with your legal team about disclaimers, ask them how it should be disclosed to stakeholders and clients, and reach out to service partners that do AI fact-checking.

In addition to potential lawsuits over AI errors, the misuse of AI tools can also diminish a company’s reputation, Zack said. “The reputational harm could hurt the company’s performance going forward,” he said. “Clients might not want to engage with them in the future because of it. Why would the next client down the road trust this company if they are in the press using AI incorrectly?”