ChatGPT, an Artificial Intelligence (AI) chatbot created by OpenAI, has taken the legal world by storm since it first launched in November 2022. The chatbot gained notoriety for its cutting-edge ability to answer questions and compose written content in a lifelike manner. Recently, however, the pitfalls associated with the use of ChatGPT in legal practice have emerged.
In June, two attorneys found themselves the subjects of a sanctions hearing in the Southern District of New York after the pair submitted a legal brief citing nonexistent case law. The author of the brief used ChatGPT to identify cases in support of his client’s position. After ChatGPT delivered seemingly spot-on cases, he proceeded to cite the AI-generated case law without reading the cases themselves or otherwise independently verifying the facts or holdings. That put the lawyer and his firm in the crosshairs of Judge P. Kevin Castel, who was none too pleased to be reading what he described as bogus judicial decisions with bogus quotes and bogus internal citations–ChatGPT had completely made up the cases. When asked about the cited cases, the attorneys initially doubled down on their mistake–continuing to stand by the fake opinions–before finally being forced to acknowledge the issue and make a formal apology to the court.
Despite claiming that they were unaware that ChatGPT could generate fictitious cases, the court fined the duo $5,000 each after finding they acted in bad faith by consciously making false statements to the court. The second lawyer, who did not conduct any legal research but whose name appeared on the brief, did not escape sanctions. The court’s imposition of sanctions against both attorneys was seemingly intended to send a strong message that ignorance is not bliss when attorneys rely on artificial intelligence to deliver legal services.
The Insurance Coverage Implications of AI-related Claims
Professional liability–also known as malpractice–insurance policies provide coverage for losses associated with the rendering of or failure to render “professional services.” In the legal context, typically “professional services” is defined to include services, including pro bono services, performed for others by the insured in their capacity as a lawyer. Thus, in the context of a claim arising from ChatGPT, the first question is whether relying on artificial intelligence to do the work of finding the case law will satisfy the definition of professional services. This should not be a close call, and insurers would be “out on a thin reed” to the extent that they try to deny coverage based on an absence of professional services because, clearly, arguments can be made that the attorneys in the claim scenario above both rendered professional services by using ChatGPT to perform the search for case law to cite in their brief–and failed to render professional services by not performing appropriate diligence to verify that the case law provided by ChatGPT was “good law.”
The much greater challenges to overcome on the insurance coverage front for this claim scenario lie in the definition of “losses” or “damages” as well as certain exclusions that address intentional conduct. As is always the case when it comes to insurance coverage, the devil will be in the details of the precise policy language under consideration when a particular claim is presented. This is especially true when it comes to the definition of “losses” or “damages.” Some policies will include definitions that are quite broad and cover any monetary amounts that an insured is required to pay to a third party, including a court and potentially opposing counsel in a fee-shifting situation. However, other policies may have a seemingly broad definition of “losses” or “damages” that is subject to exceptions such as excluding coverage for sanctions imposed against insured attorneys. Therefore, careful analysis of the dollars being paid and how that intersects with the defined terms “losses” or “damages” is an essential part of any coverage analysis.
In addition, all professional liability policies will include exclusionary language for intentional acts. But again, it is crucially important to review the precise wording of the exclusion itself before accepting any coverage denial from an insurer. In many instances, the intentional acts exclusion cannot apply unless there is deliberate fraud. The exclusion also may not apply at all to defense costs that are incurred to respond to a claim of intentional misconduct, and the exclusion may remain untriggered if a claim is resolved without an adjudication of the alleged misconduct. It is also important to consider whether the intentional conduct exclusion can be broadly applied to the firm and all its lawyers or whether there is a severability provision that maintains coverage for innocent insureds.
Considering the intentional conduct exclusion through the lens of the claim scenario discussed above, there is plenty of gray space. An argument can be made that the intentional conduct exclusion could not/should not be implicated by the mere fact that ChatGPT was used to populate the legal citations contained in the brief because this is a newer technology that the legal industry is trying to adopt to deliver more-efficient legal services to clients. While certainly not a best practice, an argument can be made that it was merely negligent for the author of the brief to rely on the accuracy of the information provided by ChatGPT, and if the attorney immediately acknowledged the mistake, an insurer would be hard pressed to raise the intentional conduct exclusion to avoid coverage for this kind of a claim.
The closer call arises when the court is made aware of the “bogus” cases and the author of the brief decides to double down on his reliance on the case law rather than come clean about the circumstances of their inclusion in the brief–and in this particular fact pattern, that is also likely the impetus for the court’s decision to impose sanctions against the lawyer. The grayest area in this claim scenario relates to the attorney whose name appeared on the brief but who was not actively involved in making use of the artificial intelligence. While an insurer would likely want to broadly trigger the exclusion for both attorneys once the matter rose to the level of sanctions, a much closer analysis of the facts related to this claim would be required to determine at what point in time the second attorney learned about the failed use of ChatGPT and whether he had direct input into the continued reliance on the case law after it was revealed to be invalid.
Conclusion
ChatGPT and other emerging artificial intelligence tools are surely here to stay–both in the legal industry and many other industries as well–because, when used responsibly, they can streamline the delivery of professional services and therefore reduce costs. However, as this recent case amply demonstrates, professionals must proceed with extreme caution and understand some of the coverage potholes that may lie ahead when a claim arises from the use of artificial intelligence technology. Insurers will surely be keeping an eye on the risks and claim activity that will flow from these tools as well, as additional exclusions and other coverage limitations may be on the horizon. Therefore, policyholders must remain ever diligent–during both the insurance policy placement phase and the claim process–to carefully review the words of their insurance policies so that maximum coverage can be accessed when the need arises.
