Early last year, when the mainstream chatter about the capabilities of ChatGPT was really starting to get going, I set out to see for myself what the generative artificial intelligence tool knew about what really matters.
I asked the chat bot some things about the ins and outs of liability insurance coverage. I was very careful in the inquiries I posed. I asked yes-no questions, ones that have definitive answers, dictated by long-standing state high court precedent. I did not ask trick questions. Nor address situations where the law is not clear or the answer depends on the particular policy language in play. In other words, I presented questions where it would be easy to measure the AI gizmo’s coverage acumen.
ChatGPT flunked the test. And not just by a little. A lot. It got 4 out of 5 questions breathtakingly wrong. If the answer should have been X, the AI program answered Y. It made my cousin Vinny look like Cardozo.
However, despite its poor showing on my narrowly-focused questions – what is the law on this particular issue, in this specific state? – the AI device was quite adept at discussing a coverage issue in broad strokes, such as, if an insured being defended under a reservation of rights is entitled to independent counsel or how to determine number of occurrences under a liability insurance policy.
[I reported my results at the time in an article that appeared in Law360 and here in Coverage Opinions.]
I decided the time had come to give ChatGPT a re-test. [I’m in the midst of grading finals for my Temple Law School insurance students. So no doubt that’s why this came to me now.]
But, of course, the test would only be fair if ChatGPT were entitled to take advantage of the technological advances it has made subsequent to its dismal performance last year. Since then, a newer version of ChatGPT is available. I shelled out $20 to get ChatGPT Plus (per month). So, instead of the freebie version, the test would be taken using the one that requires you to shell out some dough.
Once again, I asked the same type of yes-no questions that make it easy to score the technology’s coverage proficiency. I asked a few of the same questions as last year and added some new ones.
Well, ChatGPT has clearly been boning-up on insurance coverage. Instead of the “F” it got last year – which sent it to AI summer school -- the artificial intelligence apparatus earned a “C.” The improvement was remarkable.
With some exceptions, the Chatster provided correct answers. And this time, unlike last year, the technology resorted to citing case law to support its responses. And most of the time it knew the seminal decision on the issue.
Its answers were also significantly longer this go around. But therein lies the problem. That wasn’t always a good thing. By providing further explanation, the chat bot at times introduced collateral issues and rules that added vagueness and uncertainty to what could have been a much more definitive answer.
In other words, it took a correct answer for the specific state at issue and added a discussion of overarching principles and nuances and exceptions that may be generally correct – but not relevant for answering the state-specific question at issue. So, what could have been a clear answer left the reader wondering and uncertain.
Basically, it didn’t seem to know when to zip it. I guess it’s called ChatGPT for a reason.
Let’s go over the exam, which asked ChatGPT to address the following issues, all of which should be insurance coverage softballs:
- The Pollution Exclusion In Illinois
- Pennsylvania Law And The Duty to Defend
- Pennsylvania Law, Faulty Workmanship And The “Occurrence” Issue
- New Jersey Law And The Continuous Trigger
- Washington Law And Pre-tender Defense Costs
- Kansas Law And “Any Insured”
- Mystery Extra Credit Question
At the end – I know, it’s really long -- I provide my conclusion and briefly give my take on the consequences for an insurer, that used ChatGPT to make a coverage determination, and got it wrong. I addressed this issue in much more detail in last year’s article and summarize it below.
Revisiting Last Year’s Questions
The Pollution Exclusion In Illinois
In the 2023 exam, I asked ChatGPT whether, under Illinois law, the pollution exclusion precludes coverage for carbon monoxide poisoning?
It responded: “Yes, the pollution exclusion in Illinois law typically precludes coverage for carbon monoxide poisoning.”
That answer could not have been more wrong. I noted that, in Am. States Ins. Co. v. Koloms (1997), the Illinois Supreme Court said just the opposite.
This time, its answer was dramatically different [Bold text throughout this article is verbatim from ChatGPT.] The AI program stated:
Under Illinois law, the applicability of the pollution exclusion to carbon monoxide poisoning depends on the specific language of the insurance policy and the circumstances of the case. Generally, the pollution exclusion is intended to preclude coverage for damages caused by pollutants, which are typically defined as any solid, liquid, gaseous, or thermal irritant or contaminant, including smoke, vapor, soot, fumes, acids, alkalis, chemicals, and waste.
However, courts in Illinois and other jurisdictions have varied in their interpretations of whether carbon monoxide falls under the definition of a pollutant within the meaning of the exclusion. Some key considerations include:
Policy Language: The precise wording of the pollution exclusion in the insurance policy is critical. Some policies may have broad definitions of pollutants, while others may be more specific or limited.
Case Law: Illinois courts have addressed the pollution exclusion in various contexts. In some cases, courts have found that carbon monoxide does not fall within the pollution exclusion, particularly if the exclusion is ambiguous or if carbon monoxide is not specifically mentioned.
Context of Release: The manner in which the carbon monoxide was released and the specific circumstances of the poisoning incident can influence whether the exclusion applies. For example, if the carbon monoxide release is considered accidental and not part of the insured's regular business operations, courts may be more inclined to find coverage.
To illustrate, in cases such as American States Insurance Co. v. Koloms, 177 Ill. 2d 473 (1997), the Illinois Supreme Court held that the absolute pollution exclusion did not bar coverage for injuries resulting from the accidental release of carbon monoxide from a furnace. The court reasoned that the pollution exclusion was intended to address traditional environmental pollution and not routine commercial hazards such as carbon monoxide leaks from building fixtures.
Therefore, whether the pollution exclusion precludes coverage for carbon monoxide poisoning in Illinois depends on the specific facts and policy language at issue. It is advisable to consult with a legal expert or attorney who specializes in insurance law to analyze the details of the case and provide guidance based on current Illinois case law and statutory provisions.
Analysis of the answer:
At the outset, ChatGPT was spot on to cite Koloms as the key Illinois case addressing whether the absolute pollution exclusion bars coverage for injuries resulting from the release of carbon monoxide. It also got it right by noting that, under Koloms, the pollution exclusion was intended to address traditional environmental pollution. And, indeed, numerous cases in Illinois have held that the pollution exclusion does not preclude coverage on the basis that the hazardous substance at issue is not a traditional environmental pollution.
ChatGPT should have limited its answer to this. But it didn’t. As seen above, it added commentary that has relevance to a discussion of the pollution exclusion generally, but adds uncertainty to someone only interested in Illinois law, the pollution exclusion and carbon monoxide.
It also took what should have been a clear answer and made it less so, stating: “Illinois courts have addressed the pollution exclusion in various contexts. In some cases, courts have found that carbon monoxide does not fall within the pollution exclusion, particularly if the exclusion is ambiguous or if carbon monoxide is not specifically mentioned.” On that last point, however, I do not believe that any Illinois court has found that carbon monoxide falls within the pollution exclusion.
It also stated: “However, courts in Illinois and other jurisdictions have varied in their interpretations of whether carbon monoxide falls under the definition of a pollutant within the meaning of the exclusion.”
For sure, courts nationally “have varied in their interpretations of whether carbon monoxide falls under the definition of a pollutant within the meaning of the exclusion.” But, unlike ChatGPT, I would not say that about Illinois.
***
Pennsylvania Law And The Duty to Defend
Last year I asked ChatGPT whether, under Pennsylvania law, extrinsic evidence is permitted to determine an insurer’s duty to defend?
The AI tool responded: “Yes, extrinsic evidence is permitted to determine an insurer’s duty to defend in Pennsylvania.” It went on to further explain that “[i]n order to determine whether an insurer has a duty to defend, courts in Pennsylvania will review the language of the insurance policy and the facts of the underlying case, including any extrinsic evidence that is offered.”
BZZZZZT! Wrong answer. The Pennsylvania Supreme Court, in Kvaerner Metals Div. of Kvaerner U.S., Inc. v. Commercial Union Ins. Co. (2006) -- and too many other cases to count -- have said that, under Pennsylvania law, extrinsic evidence is not permitted to determine an insurer’s duty to defend.
What about now? The AI technology had this to say:
Under Pennsylvania law, the determination of an insurer’s duty to defend is generally based on the “four corners” rule, which means that courts look solely at the allegations within the four corners of the complaint and the terms of the insurance policy. This approach is also known as the “eight corners” rule, considering the four corners of the complaint and the four corners of the policy. Extrinsic evidence is typically not considered in determining the insurer's duty to defend.
However, there are some exceptions and nuances:
Clear and Unambiguous Policy Language: If the policy language is clear and unambiguous, courts in Pennsylvania will not look beyond the policy and the complaint to determine the duty to defend.
Potential Coverage: If the complaint alleges facts that potentially fall within the policy’s coverage, the insurer has a duty to defend, even if the allegations are groundless, false, or fraudulent.
Known Facts Exception: In some limited circumstances, courts may consider facts known to the insurer that are not included in the complaint but that clearly establish or negate coverage. This is not the norm and is typically applied in cases where ignoring such facts would lead to an absurd result or where the complaint itself is clearly deficient in stating the factual context.
The Pennsylvania Supreme Court case Kvaerner Metals Division of Kvaerner U.S., Inc. v. Commercial Union Insurance Co., 908 A.2d 888 (Pa. 2006), reinforced the strict application of the “four corners” rule by ruling that the insurer’s duty to defend must be determined by comparing the allegations of the complaint to the terms of the policy without resorting to extrinsic evidence.
In summary, under Pennsylvania law, extrinsic evidence is generally not permitted to determine an insurer’s duty to defend, adhering to the “four corners” rule. Exceptions are rare and typically involve situations where known facts incontrovertibly establish or negate the duty to defend.
Analysis of the answer:
ChatGPT gets it right when it said: “The Pennsylvania Supreme Court case Kvaerner Metals Division of Kvaerner U.S., Inc. v. Commercial Union Insurance Co., 908 A.2d 888 (Pa. 2006), reinforced the strict application of the ‘four corners’ rule by ruling that the insurer’s duty to defend must be determined by comparing the allegations of the complaint to the terms of the policy without resorting to extrinsic evidence.”
That’s where it should have stopped. But, alas, it didn’t, and included the following: “If the policy language is clear and unambiguous, courts in Pennsylvania will not look beyond the policy and the complaint to determine the duty to defend.”
This suggests that if the policy language is unclear and ambiguous, courts in Pennsylvania will look beyond the policy and the complaint to determine the duty to defend. That is not an exception to Pennsylvania’s strict “four corners” standard for determining an insurer’s duty to defend.
ChatGPT also stated that “courts may consider facts known to the insurer that are not included in the complaint but that clearly establish or negate coverage.”
While some “four corners” states do have such a “known facts” exception – especially when the known facts are not relevant to a coverage issue – I am not aware of any Pennsylvania decision putting Pennsylvania in that camp. Even the Third Circuit’s decision in Ramara, Inc. v. Westfield Insurance Co. (2016), which allowed, for duty to defend purposes, consideration of the fact that the worker’s compensation bar was the reason why an injured individual did not name his employer as a defendant, noted that, it doing so, it was not departing from Pennsylvania’s “four corners” rule.
New Questions Posed
I posed the following new questions to ChatGPT to see how it would do using ChatGPT Plus.
Pennsylvania Law, Faulty Workmanship And The “Occurrence” Issue
Under Pennsylvania law, is faulty workmanship an “occurrence?”
Response:
Under Pennsylvania law, the question of whether faulty workmanship constitutes an “occurrence” under a commercial general liability (CGL) policy has been addressed by the Pennsylvania Supreme Court.
In the key case of Kvaerner Metals Division of Kvaerner U.S., Inc. v. Commercial Union Insurance Co., 908 A.2d 888 (Pa. 2006), the Pennsylvania Supreme Court held that claims of faulty workmanship do not constitute an “occurrence” as defined under a standard CGL policy. The court reasoned that an “occurrence” is typically defined as an accident, including continuous or repeated exposure to substantially the same general harmful conditions, and that faulty workmanship does not fit this definition because it is not a fortuitous event.
The court in Kvaerner stated:
“The definition of ‘accident’ required to establish an ‘occurrence’ under the policy cannot be satisfied by claims based upon faulty workmanship, which do not present the degree of fortuity contemplated by the ordinary definition of ‘accident.’”
Subsequent cases have generally followed this precedent, reinforcing that purely faulty workmanship, which leads to damages that are the natural and foreseeable result of the faulty work, is not considered an “occurrence.”
However, there is an important distinction: while faulty workmanship itself is not considered an “occurrence,” there may be coverage if the faulty workmanship leads to an unintended and unexpected event causing property damage or bodily injury. For example, if faulty workmanship results in a sudden collapse of a structure causing damage to other property or injury to individuals, that resultant damage could potentially be considered an “occurrence.”
For further reading, the case Millers Capital Ins. Co. v. Gambone Bros. Dev. Co., Inc., 941 A.2d 706 (Pa. Super. 2007), also supports the principle established in Kvaerner, emphasizing that the natural and foreseeable consequences of faulty workmanship are not covered as “occurrences.”
In summary, under Pennsylvania law, faulty workmanship itself is not considered an “occurrence” under a CGL policy, but subsequent damage caused by an unintended and unexpected event resulting from the faulty workmanship may potentially be covered.
Analysis of the answer:
Once again, you guessed it, ChatGPT had it spot on, until it didn’t. Unquestionably, Kvaerner and Gambone are the two leading cases in Pennsylvania on the issue whether faulty workmanship is an “occurrence.”
And the AI device correctly stated their holdings. Under Kvaerner, the definition of “accident” required to establish an “occurrence” under the policy cannot be satisfied by claims based upon faulty workmanship. Likewise, it knew that, under Gambone, the natural and foreseeable consequences of faulty workmanship are not covered as “occurrences.”
But it marched on, also noting that, “while faulty workmanship itself is not considered an ‘occurrence,’ there may be coverage if the faulty workmanship leads to an unintended and unexpected event causing property damage or bodily injury.”
When it comes to coverage for faulty workmanship, that is certainly a correct statement of the law, generally, in many states. However, it is not correct under Pennsylvania law, where any potential coverage for the consequential damages of faulty workmanship is governed by the specific language in Gambone and how the issue was addressed by the Pennsylvania Superior Court in its 2013 decision in Indalex, Inc. v. National Union.
***
New Jersey Law And The Continuous Trigger
What is the trigger of coverage in New Jersey for latent bodily injury or property damage?
Response:
In New Jersey, the trigger of coverage for latent bodily injury or property damage under a standard commercial general liability (CGL) policy typically involves a combination of judicial doctrines and interpretations, often referred to as the “continuous trigger” theory. This approach is particularly relevant for claims involving latent injuries or damages that develop over an extended period.
Here are the key aspects of the trigger of coverage in New Jersey for latent bodily injury or property damage:
Continuous Trigger Theory: New Jersey courts have adopted the continuous trigger theory in cases involving long-term exposure to harmful conditions, such as environmental contamination, asbestos-related injuries, and other latent injuries. Under this theory, coverage is triggered continuously from the time the exposure begins until the manifestation of the injury or damage. This means multiple insurance policies across different policy periods can be triggered.
Key Case - Owens-Illinois, Inc. v. United Ins. Co.: The New Jersey Supreme Court’s decision in Owens-Illinois, Inc. v. United Ins. Co., 138 N.J. 437 (1994), is pivotal in establishing the continuous trigger theory. The court held that for long-tail claims, such as those involving asbestos exposure, all policies in effect from the time of first exposure to the harmful condition through the manifestation of the disease or injury are triggered.
Allocation of Damages: In cases involving multiple triggered policies, New Jersey courts have also adopted the “pro rata by time on the risk” allocation method. This approach allocates damages among all triggered policies based on the time each policy was in effect relative to the entire period of injury or damage development.
Other Key Decisions: Other important cases that have shaped the trigger of coverage in New Jersey include Carter-Wallace, Inc. v. Admiral Insurance Co., 154 N.J. 312 (1998), where the court applied the continuous trigger theory to environmental contamination claims, and Spaulding Composites Co. v. Aetna Casualty & Surety Co., 176 N.J. 25 (2003), which reinforced the application of continuous trigger in long-tail claims.
Implications for Insureds and Insurers: The continuous trigger theory ensures that insureds have access to coverage across multiple policy periods, which can be critical in addressing the substantial costs associated with latent injuries and damages. For insurers, it emphasizes the importance of understanding potential long-term exposure and liability across policy periods.
In summary, the trigger of coverage for latent bodily injury or property damage in New Jersey typically follows the continuous trigger theory, as established by key court decisions. This approach involves the activation of coverage from the time of initial exposure to the harmful condition through the manifestation of the injury or damage, with damages allocated among all triggered policies based on their time on the risk.
Analysis of the answer:
While it was a little redundant, it is hard to argue that ChatGPT did not accurately describe the trigger of coverage in New Jersey for latent bodily injury and property damage. It knew that it’s the continuous trigger, how the trigger method is applied, that Owens-Illinois is the seminal case and the impact that a continuous trigger has on the availability of coverage is applicable claims scenarios.
However, when it decided to show off, by also mentioning allocation, things got off track. Yes, New Jersey courts employ pro-rata allocation, but taking into account both time on the risk and policy limits. While Chat GPT’s statement, that New Jersey courts apply “pro rata by time on the risk” allocation was not correct, it at least found the right camp, i.e., not stating that it is “all sums.”
I decided to see if it fared any better when specifically asked about allocation under New Jersey law. It did not, once again answering that it is simply “pro-rata time on the risk.”
***
Washington Law And Pre-tender Defense Costs
Under Washington law, is an insurer liable for pre-tender defense costs?
Response:
Under Washington law, insurers generally are not liable for pre-tender defense costs. This principle is based on the idea that an insurer’s duty to defend is triggered by the tender of the claim or suit to the insurer. Therefore, any defense costs incurred before the insured notifies the insurer (i.e., pre-tender costs) typically are not covered.
This approach was affirmed in the case of Weyerhaeuser Co. v. Aetna Casualty & Surety Co., 123 Wash. 2d 891, 874 P.2d 142 (1994), where the Washington Supreme Court held that an insurer is not responsible for defense costs incurred before the insured tenders the defense of the action to the insurer. The court reasoned that the duty to defend arises at the point of tender, and costs incurred prior to that tender are not the insurer’s responsibility.
However, there are exceptions and nuances to consider:
Waiver or Estoppel: If an insurer’s actions cause the insured to reasonably incur defense costs before tendering the claim, the insurer might be estopped from denying those costs. For example, if the insurer delays in responding to a notice of a claim in a way that prejudices the insured, it might be responsible for some pre-tender costs.
Policy Language: The specific terms of the insurance policy can affect the outcome. If the policy explicitly provides coverage for pre-tender costs, those provisions would govern.
Timely Notice Requirement: The insured is generally required to give timely notice of the claim to the insurer. Failure to do so might absolve the insurer of the responsibility for pre-tender costs, unless the insurer cannot demonstrate that it was prejudiced by the delay.
Overall, while the general rule in Washington is that pre-tender defense costs are not covered, the specifics of each case and the terms of the insurance policy can lead to different results.
Analysis of the answer:
Big goof here for the supposed know it all. In 2013, the Washington Supreme Court held in National Surety Corp. v. Immunex Corp. that “the duty to defend arises not at the moment of tender, but upon the filing of a complaint alleging facts that could potentially require coverage. . . . Accordingly, an insured can recover pretender fees and costs except where a late tender prejudiced the insurer.”
Unlike with other questions, ChatGPT missed the seminal case on the issue. Not to mention, the supposed case it cited was decided twenty years earlier and reached the opposite conclusion. Well, at least there’s one place where Washington law is not pro-policyholder.
***
Kansas Law And “Any Insured”
This was the hardest question on the exam.
Under Kansas law, is coverage owed under the liability section of a homeowner’s policy to a husband for negligence that contributed to his wife’s killing of a child in their care when a policy exclusion applies to “bodily injury which is expected or intended by any insured?”
Response:
Under Kansas law, the interpretation of insurance policy exclusions, particularly those excluding coverage for “bodily injury which is expected or intended by any insured,” often depends on the specific facts of the case and the precise wording of the policy.
Here are key points to consider:
Policy Exclusion Language: The exclusion for “bodily injury which is expected or intended by any insured” generally aims to preclude coverage for intentional acts. The critical issue is whether the husband’s alleged negligence can be separated from the intentional act of the wife.
Joint Obligations of Insureds: Many homeowner policies include exclusions that apply to acts of “any insured.” If the policy states that the exclusion applies to any insured, the intentional act of one insured (the wife, in this case) may preclude coverage for all insureds under the policy, including the husband.
Case Law Interpretation: Kansas courts have addressed similar exclusions. In Bell v. Tilton, 234 Kan. 461 (1983), the Kansas Supreme Court held that where an exclusion applies to acts “expected or intended by the insured,” the subjective intent of the insured who committed the act is critical. However, when the policy states “any insured,” the exclusion can apply even if another insured did not intend or expect the injury.
Negligence vs. Intentional Acts: If the husband’s liability is based purely on negligence (e.g., failing to supervise), some courts may differentiate between the negligent conduct and the intentional act. However, if the policy’s exclusion is broad and applies to any insured, the exclusion could be interpreted to bar coverage for the husband’s negligent acts connected to the wife’s intentional act.
In the context provided, if the policy exclusion is broad and applies to bodily injury expected or intended by “any insured,” Kansas law is likely to uphold the exclusion, precluding coverage for the husband’s negligence related to the wife’s intentional killing of the child. The rationale is that the intentional act of one insured (the wife) triggers the exclusion, affecting all insureds under the policy.
Therefore, based on the exclusion for “bodily injury which is expected or intended by any insured,” it is likely that coverage would be denied under Kansas law for the husband’s negligence in connection to his wife’s intentional act. However, precise outcomes can vary based on specific policy language and case-specific facts, so consulting a legal expert familiar with Kansas insurance law is advisable for a definitive interpretation.
Analysis of the answer:
This answer is not correct. The question posed was taken from the Kansas Supreme Court’s 1998 decision in Brumley v. Lee, where the court examined an exclusion for “bodily injury … which is expected or intended by any insured” and held that coverage was owed to an insured-husband for negligence that contributed to his insured-wife’s killing of a child in their care. The court held that the policy’s severability clause created an ambiguity that required a construction that favored the insured.
ChatGPT was generally correct that “[m]any homeowner policies include exclusions that apply to acts of ‘any insured.’ If the policy states that the exclusion applies to any insured, the intentional act of one insured (the wife, in this case) may preclude coverage for all insureds under the policy, including the husband.”
However, the AI device did not mention that, under Kansas law (as in some other states), such interpretation of “any insured” is altered by a policy’s severability clause.
What’s more, the case cited by ChatGPT, Bell v. Tilton, is simply not an “any” insured case. It was a straightforward BB gun shooting --“expected or intended” case, concluding that the exclusion applied “if, from the acts, circumstances, and inferences of the case, it appeared [the insured] had the desire to cause the consequences of his acts or he believed the consequences were substantially certain to result[.]”
Extra Credit
I have always said that Elvis loved to tender to his insurance company. I asked ChatGPT Plus if that’s true?
Response:
There is no public record or well-known anecdote suggesting that Elvis Presley had a particular fondness for tendering claims to his insurance company. Elvis Presley, like many celebrities, undoubtedly had insurance policies to protect his assets, health, and career, but his personal feelings or habits regarding insurance claims are not documented in the public domain.
Elvis’s public persona and legacy are more focused on his music, movies, and cultural impact rather than mundane aspects of life like dealing with insurance. If you're looking for information on how celebrities typically manage their insurance, they often have agents and managers who handle such matters to ensure their valuable assets and well-being are protected.
Huh? What are you calling “mundane?”
ChatGPT Plus’s Grade
It was a mixed bag for ChatGPT Plus.
Some of its answers were right on, including citing the seminal decision(s) on the issue.
Others were correct, until it decided to chat further, and discuss overarching principles that may be correct generally in some states, but not the one at issue or relevant to answering the question. By adding such additional explanation, the chat bot at times introduced collateral issues that added vagueness and uncertainty to what could have been a much more definitive answer.
And some of its answers left a lot to be desired.
For sure, ChatGPT Plus is much improved from last year’s “F.” Of course, it had only one direction to go. I give it a “C.”
If you need to know enough about a coverage issue to discuss it at a cocktail party, you're pretty safe sneaking a peek at ChatGPT. And, if the person you are speaking to does not know the answer, you are in even better shape.
But using ChatGPT (even the 20 bucks a month version) to address coverage in a more serious setting, especially to make a coverage determination, would be foolhardy.
Also important to keep in mind -- I tested ChatGPT’s aptitude for coverage based solely on case law. Sure, in some cases, coverage will rise and fall based principally on case law. The policy language, whatever the version, is not going to alter the outcome.
But, of course, in many others, policy language can make a significant difference. This is where I see ChatGPT’s biggest weakness in making coverage determinations. Even if it eventually gets the case law down cold, the specific policy language – and the fact that the entire policy can be relevant to interpreting a single provision -- is surely going to be a challenge for AI to overcome.
No doubt artificial intelligence, like most technologies, will improve as time goes on. And it certainly has. But, for now, and probably a long time, despite the praise being heaped on AI, insurers would be well-served to resist the temptation to use a coverage lawyer in a chat box.
ChatGPT: The Consequences for Claims Handling
What are the potential consequences for an insurer, that used ChatGPT to make a coverage determination, and got it wrong? I addressed this in further detail in last year’s article. Here’s a summary.
Courts almost always conclude that an insurer, despite getting the answer to a coverage question wrong, has not committed bad faith. The various state standards to prove bad faith are generally high – the conduct had an intentional aspect; the issue was not fairly debatable; the denial was unreasonable (and others, along these lines). Reaching an incorrect coverage determination is unlikely to satisfy this heavy burden.
However, while getting a coverage determination incorrect is unlikely to lead to bad faith, the same cannot always be said about the manner in which an insurer reached the incorrect determination. Bad faith can occur when an insurance company makes an inadequate investigation or fails to perform adequate legal research concerning a coverage issue.
An insurer that uses an artificial intelligence program to answer a legal coverage question, and then uses that answer to improperly deny coverage – especially with known flaws and limitations in the application’s accuracy – is unlikely to get the benefit of having acted reasonably for purposes of defeating a bad faith finding.
|
