AI is a broad term referring to any machine-based system that can make predictions, recommendations or decisions influencing real or virtual environments for a given set of human-defined objectives.

It refers to all manner of techniques used to search and analyze large volumes of data; robotics dealing with the conception, design, manufacture and operation of programmable machines; and algorithms and automated decision-making systems able to predict human and machine behaviour.

Lawyers have been using AI for decades: spam filters, spellchecks, search terms in electronic research and automated document generation are but a few examples.

Large language models (LLMs) are types of computer programs that focus on language processing to produce text that resembles what humans would produce. They are trained on vast amounts of text data, such as books, articles and material available on the internet, to understand human language patterns, grammar and semantics.

LLMs generate answers based on the probability of a word or combination of words that is most likely to come next. By predicting what comes next, the platform can generate human-like content that is statistically likely in response to a prompt. It can be told to mimic specific writing styles or tones, present the information in various formats and shorten or lengthen the response as directed.

Gen AI is a particular type of large language model. Unlike other forms of AI, it doesn’t simply reproduce material it finds. It generates material based on information ‘prompts’ that the user inputs, which typically consist of short instructional texts. Its answers are fluent and appear to be human-generated.

Lawyers can use Gen AI to create human-like text in tasks such as client onboarding, scheduling appointments, billing, document review, legal research and drafting. This powerful tool can be used to automate repetitive and time-consuming tasks, improve the quality of written material, and free up valuable time for higher-level analysis and strategic decision-making.

The risks and benefits can vary significantly between public tools like ChatGPT and private AI solutions. For example, Lexum is using AI across their legal research services in ‘private’ mode. A number of larger law firms have already developed in-house AI tools for research purposes by their lawyers and staff. No matter what the solution, they all require diligence and training for staff prior to use.

ChatGPT is the name of a particular Gen AI product. There are free and paid versions (GPT-3.5 and GPT-4). ChatGPT is not alone in the field, with a number of well- and lesser-known competitors including Microsoft’s Copilot, Google’s Bard and PaLM2. Companies like LawDroid, Rally, Harvey.AI and LexisNexis are also building legal profession-specific models.

This is just a small sampling of the Gen AI products currently available. More are being released daily.

OpenAI is the name of the company that created ChatGPT.

In their Notice to the Public and Legal Profession: Ensuring the Integrity of Court Submissions When Using Large Language Models, the Alberta Courts acknowledged that emerging technologies like Gen AI bring both opportunities and challenges, and that the legal community must adapt accordingly.

Their Notice urged litigants to exercise caution when citing legal authorities or analysis derived from LLMs and emphasized that “it is essential that parties rely exclusively on authoritative sources such as official court websites, commonly referenced commercial publishers, or well-established public services such as CanLII” for any references to case law, statutes or commentary in representations to the courts.

The Courts called for “humans in the loop” and stipulated that all AI-generated submissions must be verified with “meaningful human control” that cross-references reliable legal databases to ensure that citations and their content hold up to scrutiny.

This obligation accords with the longstanding practice of lawyers to verify their sources.

The Federal Court “recognize[d] that emerging technologies often bring both opportunities and challenges” but went further than the Alberta courts and required litigants to inform the court and other parties “if they have used artificial intelligence to create or generate new content in preparing a document filed with the Court.” This must be done in writing in the first paragraph of any such document.

As with the courts of Alberta, the Federal Court’s Notice to the Parties and the Profession: The Use of Artificial Intelligence in Court Proceedings “urges caution when using legal references or analysis created or generated by AI, in documents submitted to the Court. When referring to jurisprudence, statutes, policies, or commentaries in documents … it is crucial to use only well-recognized and reliable sources. These include official court websites, commonly referenced commercial publishers, or trusted public services such as CanLII.”

Humans must once again be “in the loop.” The Federal Court note states that it is “essential to check documents and material generated by AI. The Court urges verification of any AI-created content in these documents.”

The risk of inadvertent disclosure of confidential client or proprietary information cannot be overstated. Gen AI platforms like ChatGPT were not built with confidentiality in mind, so information contained in prompts may easily find its way into the public domain if care is not taken at all times. Our ethical rule of confidentiality, Rule 3.3 of the Code, is broad and requires continuing due diligence to hold in strict confidence all information concerning the business and affairs of a client throughout all aspects of a client relationship and beyond.

For example, in one report involving Samsung, engineers used ChatGPT to help fix problems with their source code. In doing so, they inputted the source code itself into their prompt. They also asked ChatGPT to convert internal meeting notes into a presentation. Once entered, there was no way to retrieve or delete the compromised data. Samsung’s trade secrets were then available to Open AI, the company behind ChatGPT, for purposes completely unrelated to Samsung.

Any content you upload to a Gen AI engine, whether public or private, should be carefully considered. For example, uploading content such as a memo for refinement, or uploading a document for proofreading, means the content of that document is now available for the Gen AI engine and your content could potentially be used to train that engine and for other purposes. Uploading that content could risk privacy and you may not be able to retrieve your information once it has been uploaded.

This issue may be addressed through the use of internal or proprietary AI platforms, a number of which are currently in development, but best practice is to exercise extreme caution throughout.

Best Practices:

• Never include confidential or potentially identifying information in prompts.
• Use only non-identifiable information in prompts.
• Use caution at all times.

Particularly in the case of impersonation and fraud, Gen AI tools can be used for criminal purposes. For example, using an AI speech generator, a sound bite can be taken from a voice mailbox greeting to impersonate that person. As with any powerful tool, the tool can become a weapon, so caution is needed.

The free version of ChatGPT-3.5 has a knowledge cutoff date of September 2021. This means that events after this time will not be reflected in its responses.

ChatGPT-4’s cutoff has been updated to November 2023 so it is more recent but will still not be aware of current events in all cases. This may affect the reliability of its responses as well.

As a ‘generative’ tool, Gen AI is designed to generate words or images based on the information it encounters in its training. This is not a flaw. By design, it fabricates information when it doesn’t have sufficient data to answer a prompt. These events are known as ‘hallucinations.’

In the legal context, this is highly problematic. Gen AI tools are not tied to a foundation of truth or reality and are designed to provide creative responses to queries.

Even though these products can assist in retrieving relevant case law, statutes and legal articles from vast databases, there are notorious examples of how they have filled gaps in data by making up names, dates, historical events and even legal cases. If pressed, they may even write the cases themselves.

Recognizing that this happens, any lawyer using Gen AI for substantive legal work must proceed with caution and ensure that they independently verify all information generated by the platform. Lawyers should never rely on Gen AI to judge its own accuracy.

Gen AI models are trained on data from the internet. If the data is unreliable, so too will be the resulting output.

As a result, Gen AI tools can produce content that is discriminatory, biased or that reinforces stereotypes.

For example, a Bloomberg study evaluating a Gen AI image generator found that images created of high-paying jobs were dominated by subjects with lighter skin tones, while subjects with darker skin tones were more commonly generated by prompts like “fast-food worker” and “social worker.”

Similar results were found for prompts based on by gender. Most occupations in the dataset were dominated by men, except for lower-paying jobs like housekeeper and cashier. Men with lighter skin tones represented the majority of subjects in every high-paying job, including “politician,” “lawyer,” “judge” and “CEO.”

Part of the problem is the opacity of Gen AI models. Because Gen AI operates as a black box, it is difficult to assess the validity of the inputs it relies on or trace how the system produces its outputs.

If lawyers cannot assess Gen AI systems for bias, they may inadvertently perpetuate stereotypes, violate human rights legislation and damage public trust in the justice system.

Best Practices:

• Review generated content to ensure that it aligns with your ethical and legal obligations. This includes assessing for biases or stereotypical associations.
• Use prompts that minimize biases.

While it may seem like Gen AI tools create new material from independent thought processes, that is not how they function. The platforms are trained on data scraped from the internet — billions of data points that can include copyright-protected materials like articles, books, code, paintings or music — that the platforms use to identify patterns and use to respond to a prompt.

If the AI-generated outputs contain material that is identical or substantially similar to copyright-protected work, there is a risk that they infringe the original copyrights. At this juncture, it is still unclear who owns content created by Gen AI platforms. Lawyers need to understand the risks this represents and how to protect themselves. They need to ask their providers whether their models were trained with any restricted content.

Going forward, new Gen AI platforms may be developed that train themselves on internal or proprietary datasets rather than the internet generally. This would avoid the risk of copyright infringement, let the platforms produce content in the same style as their users’ previous work, and create a more transparent audit trail.

Best Practice:

• Disclose the use of Gen AI any time it is relied upon.

Client Satisfaction: Some consumers of legal services will be uncomfortable with the idea that their legal work is performed by a machine. Others will appreciate the speed and convenience that AI platforms offer. Firms that have determined they can safely use Gen AI can, if optimized correctly, respond to client comments, questions and concerns immediately, at all hours.

Automated Client Intake and other Administrative Tasks: AI chatbots can automate client onboarding. Routine documents can be turned into smart forms that clients complete with AI assistance. Chatbots can be used to help clients navigate the law firm’s website and fill out questionnaires, which are then automatically entered into the firm’s client management software. Meetings can be scheduled quickly and easily, and calendars can be automatically updated.

Live Chat: For lawyers wanting to add chat features to their websites, Gen AI can help them accomplish that goal. While basic chatbots can answer yes/no questions and provide FAQs, Gen AI can be used to respond to more complex questions and develop much more friendly and seamless client interactions. As a precaution, the chatbot should disclose that it is not providing legal advice generated by a lawyer.

Brainstorming: Some lawyers are using Gen AI programs to prepare for trial. This can include everything from developing questions for potential jurors during jury selection to explaining expert reports to developing arguments and determining the order in which arguments should be made.

Preparing Cross-Examination: By listing the facts and asking for suggestions on how to approach them, Gen AI can help structure a questioning or cross-examination.

Argument Validation and Counter Argument Consideration: Lawyers can ask Gen AI to assess the strengths and weaknesses of their arguments. It can be asked to anticipate opposing counsel’s counter arguments.

Create Agendas, Memos and Contracts: Gen AI can be used to automate the process of generating documents. By giving the tool specific input and requirements, lawyers can use Gen AI to create meeting agendas and agreements. It can be told to prepare a memo instructing an associate to research a point or draft a document. It can take the resulting research memo and turn it into a brief. Note, however, that free versions of some tools struggle to produce quality material. More advanced, paid models may be more reliable, and some firms are developing tools that may be even more reliable because they are trained on legal data rather than what may be generally available on the internet.

Drafting Correspondence, Proposals and Outlines: Lawyers can use Gen AI to improve their correspondence, project proposals, cover letters, resumes, profiles and other written communications. By analyzing the context and purpose of the communication, they can serve as ‘creative collaboration’ tools that generate quality professional text. They can re-write your letters to fit your client’s style. They can suggest alternate phrasing, provide example of more persuasive arguments, and ensure that documents conform to drafting conventions and requirements. Gen AI is particularly useful to generate ideas and assist in writing a first draft or outline.

Identifying Risks and Inconsistencies in Contracts: Gen AI can analyze documents to identify potential risks, ambiguities and inconsistencies. By automating repetitive tasks, such as identifying specific clauses, checking cross-references or flagging anomalies, the tool can accelerate the review process, saving time and resources. Lawyers can streamline their document review and focus instead on critical issues, compliance and risk mitigation.

Creating Presentations: Because Gen AI offers excellent summarizing skills and can incorporate written words and images, it can be used to prepare a PowerPoint presentation based on memos or articles the lawyer has already written.

Gen AI can be a valuable assistant that summarizes lengthy legal documents, precedents, legal briefs, expert reports or medical records. It can break down legal jargon and complicated issues. It can sort through large volumes of data and identify key documents for a case. It can distill complex information into concise and understandable summaries.

Potential Time and Cost Savings: One of the major benefits of Gen AI is rapid idea brainstorming and drafting.

Lawyers who learn to use Gen AI safely and responsibly will likely see a competitive advantage. Time currently spent filling out forms, drafting emails and advertising can be reduced. Increasing the efficiency of firm operations can lower costs to the client and make access to professional legal advice more affordable, thus increasing access to justice and legal services generally.

Reduce Procrastination, Eliminate Mental Blocks: Gen AI can help reduce procrastination by providing the initial spark needed to get work done. It can overcome mental blocks interfering with a writer’s creativity. In the context of performing due diligence, case preparation and legal opinion drafting, it can free the lawyer to realize significant savings and efficiencies and work on higher-value tasks.

Proofreading: Gen AI can improve the wording and structure of any document and fix spelling and grammatical errors. It can recommend language improvements and ways to improve readability to make the result clearer and more convincing.

Website Content, Blog Posts, Marketing Emails, Tweets, LinkedIn Posts and Newsletter Articles: Gen AI can summarize and edit memos, reports and other material. It can translate resources to and from other languages. Since initial drafts can be created in a matter of minutes, the firm or administrative staff stand to realize significant time and cost savings. Gen AI’s uses are limited only by the lawyer’s imagination and their ethical obligations under the Code of Conduct, the Rules of the Law Society of Alberta and the Legal Profession Act.

As more advanced models are developed, and legal industry-specific tools are built on general-purpose programs like GPT-4, the capacity and accuracy of Gen AI and LLMs will continue to evolve.

Currently, there are several legal Gen AI programs available, but they may not be ready for full scale adoption. In the near future, however, these tools will exit beta testing and become mature products available to law firms of all sizes.