Published – Summer 2024 | The Litigator
In 2023, The New York Times reported on a lawyer who relied on ChatGPT to conduct legal research for a motion filed in U.S. federal court. The issue only came to light when opposing counsel was unable to locate several of the cases cited in the brief. The court ultimately determined that the authorities did not exist. The lawyer acknowledged that the research came from a source that proved to be unreliable and admitted that he did not understand the technology was capable of fabricating case law. The court found that the lawyer had acted in bad faith and imposed a $5,000 fine. At the time, it was unclear whether separate professional discipline followed.
Several months later, media reports revealed that Michael Cohen, former counsel to Donald Trump, had provided his own lawyer with fictitious legal citations generated by Google Bard. Those citations were included in a federal court filing, again resulting in public scrutiny and judicial concern.
These well-publicized examples are likely only a small sample of instances where artificial intelligence (“AI”) has been used unsuccessfully in legal practice. They raise an important question for Canadian insurers and lawyers alike: how frequently is AI being used in the insurance and litigation context, and what are the consequences? This article examines the role of AI in insurance claims handling and defence litigation, with particular attention to its potential advantages, limitations, and the evolving response of Canadian courts.
What Is Artificial Intelligence?
The term “artificial intelligence” dates back to 1956 and refers to a broad category of technologies rather than a single system or function. There is no universally accepted definition. In general, AI involves the use of algorithms, machine learning, and natural language processing. AI is commonly described as either narrow (or “weak”) AI, which performs specific tasks, or general (or “strong”) AI, which remains largely theoretical and would be capable of performing a wide range of human cognitive functions.
Most AI currently used in the insurance industry falls within the narrow AI category. These systems rely on machine learning, natural language processing (“NLP”), and computer vision. Machine learning uses algorithms trained on data sets to identify patterns, classify information, and make predictions. NLP focuses on enabling computers to interpret written and spoken language, while computer vision allows systems to extract information from images and video.
AI and Insurance Claims Handling
Insurers have begun integrating AI into various aspects of their operations, including claims intake and processing, underwriting, document review, image analysis, fraud detection, loss valuation, and risk assessment. The use of these tools may result in more consistent workflows, reduced manual error, faster decision-making informed by historical data, and earlier identification of potentially fraudulent claims.
Insurance technology, often referred to as “InsurTech,” continues to expand. Industry research has shown sustained interest in AI investment, driven by the potential to process information more efficiently and manage increasing claim volumes. As these technologies mature, they are influencing how insurers interact with policyholders and how claims are evaluated, adjusted, and litigated.
The most developed application of AI within insurance is claims automation. Claims handling has traditionally been one of the most resource-intensive components of an insurer’s business. Because many claims-related tasks are repetitive and rules-based, they are well suited to automation.
For example, AI-powered chatbots may be used during the initial reporting stage to gather information from insureds following property damage or accidents. Machine learning and NLP tools can review submitted documents to extract relevant details, while computer vision systems can assess photographs to evaluate the extent of damage. These technologies may also identify irregular patterns that warrant further scrutiny for potential fraud. In many cases, AI-generated outputs are reviewed by human adjusters before a claim is finalized, particularly where coverage issues or higher-value losses are involved.
Some insurers have publicly reported that AI-assisted claims can be reviewed within seconds. While such speed can reduce administrative delays, it also underscores the importance of appropriate oversight.
Despite potential efficiencies, AI-driven claims decisions must comply with Canadian legal and regulatory requirements. Questions remain about the reliability of AI-generated conclusions, particularly where those conclusions form the basis for coverage denials or litigation strategies.
A significant concern is the so-called “black box” problem. Many AI systems do not provide clear explanations for how decisions are reached, making it difficult to trace the reasoning process. Researchers and advocacy organizations have raised concerns that AI tools may replicate or amplify bias present in historical training data, leading to discriminatory outcomes.
Speed can also introduce risk. If an AI system relies on flawed assumptions, those assumptions may be applied repeatedly and at scale. Without appropriate safeguards, biased or inaccurate decision-making could expose insurers to allegations of improper claims handling or bad faith. Although AI-related bad faith claims have not yet been tested extensively in Canadian courts, it is foreseeable that litigants may seek disclosure of algorithms, models, or source code used to support claims decisions.
AI and Defence Litigation
The increased use of AI has prompted discussion about the future role of defence counsel. While AI tools are reshaping certain aspects of legal practice, they have not replaced the need for lawyers. Litigation remains a human-centred process that depends on judgment, credibility, and advocacy.
That said, tasks traditionally performed by lawyers—such as legal research, document review, and drafting—are changing. Predictive tools trained on case law, legislation, and regulations can now synthesize large volumes of information quickly. Commercial platforms advertise the ability to assess how courts may approach particular legal issues, and AI-assisted research can significantly reduce turnaround times.
Faster access to research may allow lawyers to provide advice more efficiently and manage costs more effectively. Insurers, who have long imposed strict time expectations for research and review, may increasingly rely on their own AI tools for these purposes. As a result, the amount of time counsel spend on certain tasks may continue to decrease.
AI can also assist with preparing chronologies, summarizing evidence, drafting pleadings, and organizing discovery materials. However, meaningful oversight by experienced lawyers remains essential. AI outputs depend heavily on the quality and completeness of the data provided. These systems cannot independently assess witness credibility, weigh competing evidence, or respond dynamically during examinations for discovery.
Litigation strategy also relies on professional judgment developed through years of interaction with judges, mediators, arbitrators, experts, and opposing counsel. These skills cannot be replicated by current AI systems. Any approach that lacks this human element risks overlooking important nuances.
Accuracy and reliability are central to legal practice. Lawyers who use AI must ensure that the tools they rely on draw from current and dependable sources. Failure to do so may result in errors that undermine both client interests and counsel’s standing before the court.
Judicial Response in Canada
Canadian courts have begun to address the use of generative AI in litigation. In Ontario, amendments to Rule 61 of the Rules of Civil Procedure now require factums filed on appeal to include a certificate confirming counsel’s satisfaction with the authenticity of the authorities cited.
Other courts have issued more explicit guidance. Practice directions from the Supreme Court of Yukon and the Court of King’s Bench of Manitoba require parties to disclose whether AI tools were used in preparing legal research or submissions, and for what purpose.
In December 2023, the Federal Court released a notice addressing the use of AI in court proceedings. The Court expects parties to inform both the Court and opposing counsel if AI was used to generate new content in filed materials. Where AI-generated content is included, that fact must be disclosed at the outset of the document. The notice reflects judicial awareness of both the potential efficiencies and the risks associated with AI, including fabricated authorities and biased outputs.
Conclusion
AI continues to develop at a rapid pace, presenting both opportunities and challenges for insurers, defence counsel, and the courts. While AI may influence how claims are processed and how litigation files are managed, it does not remove the need for careful legal analysis, ethical decision-making, and professional judgment.
As cost considerations remain a priority for insurers, lawyers will increasingly encounter AI-generated data and tools in their practices. Understanding the limitations as well as the potential uses of these technologies is essential. AI may change how legal work is performed, but it has not replaced the role of defence counsel in navigating risk, evidence, and advocacy within the Canadian legal system.
