The Illinois ARDC AI implementation guide gives lawyers a practical way to adopt generative AI without losing sight of their ethical duties. Formally titled The Illinois Attorney’s Guide to Implementing AI, it was released effective October 1, 2025. The guide was written with solo and small-firm lawyers in mind, since many work without dedicated IT support. Still, its framework applies to firms of every size.
It serves as a nonbinding companion to the Illinois Supreme Court’s Policy on Artificial Intelligence, which took effect on January 1, 2025. That policy allows lawyers to use AI, provided they continue to meet existing professional obligations. The guide provides practical details. It explains how these tools work, how to match a tool to the sensitivity of client data, and how to communicate with clients about AI use. Sample policies, forms, and a checklist are also included. For lawyers who feel pressure to adopt AI quickly, the guide offers a structured place to start.
Key Takeaways from the Illinois ARDC AI Implementation Guide
The ARDC ties its approach directly to the Supreme Court’s policy. According to the guide itself, the policy makes four key points:
It authorizes AI use.
It holds lawyers accountable for AI-assisted work product.
It requires competence with the tools.
It demands protection of private and confidential information.
The competence point reflects Illinois Supreme Court Rule 1.1, which already extends a lawyer’s duty to the benefits and risks of relevant technology. The confidentiality point corresponds to Rule 1.6 and a lawyer’s duty to safeguard client data. For help with the accountability duty, the guide points lawyers to tips and resources in the ARDC’s PMBR self-assessment program, which this overview explains further.
The guide then explains the technology in plain terms. Nearly every AI tool lawyers encounter runs on a transformer, a technology introduced in 2017. AI companies and their models are compared to car makes and models, much like Honda and its Accord. OpenAI builds GPT, for example, and Anthropic builds Claude. Many legal tech products are built on top of models from these same providers. Because client data travels to that underlying model, lawyers need to know who produces it and what safeguards protect it.
A Three-Step Framework for Choosing AI Tools
The core of the guide is a nonbinding, three-step framework that can help lawyers document reasonable safeguards under Rule 1.6:
Classify the information. The guide ranks data from lowest to highest risk: general information, de-identified information, confidential information, and sensitive personal information such as Social Security numbers, account numbers, medical records, and biometric identifiers.
Categorize the tool. Third-party managed tools include native platforms such as ChatGPT, Claude, and Gemini, along with applications that connect to those models through an API. Self-managed tools run on hardware the firm controls.
Evaluate the safeguards. Key areas include authentication, model training, data retention, data isolation, supply chain risk, risk management frameworks, security audits, and terms of use.
Third-party tools are then sorted into public, consumer, business, and enterprise categories. Public and consumer tools are considered less appropriate for confidential or sensitive information. As a general rule, confidential information should not be sent to any tool that allows it to be used for model training.
The guide also warns against trusting labels. A paid or business version is not automatically safe, and a tool that trains on data by default may still offer an opt-out. The actual safeguards, not the marketing, determine whether a tool fits the data.
Illinois ARDC AI Implementation Guide on Client Communication Safeguards
Client communication is treated by the guide as an added safeguard. It describes a notice-and-opt-out approach for most data, with stronger protections for highly sensitive information. For example, when a business-class tool processes confidential information, the guide encourages notice and a right to opt out. For sensitive personal information, it encourages enterprise-level safeguards, informed client consent, or both. Importantly, the guide states that obtaining consent cannot replace the due diligence a lawyer owes in selecting and using AI tools.
The guide also acknowledges a growing challenge: AI features are being built into platforms like Microsoft 365 in ways clients may not be able to opt out of. In those cases, enterprise-level safeguards are recommended to offset the lack of opt-out or consent rights.
To put all of this into practice, the guide includes a Practice Resource Kit built around a hypothetical small firm. It contains a terms of use checklist, a sample notice of AI practices, a sample firm policy on the use of AI tools, and a sample informed consent form. Lawyers who want broader context on these duties can review this to look at the legal ethics of AI.
Documenting Artificial Intelligence Decisions as Legal Technology Evolves
The ARDC acknowledges that AI changed almost daily while the guide was being written. For that reason, it was built around core principles rather than specific products. It already anticipates agentic AI, which can access information and take actions on its own, and the ARDC plans to update the guide regularly. Its closing advice is simple: stay curious, apply safeguards, document decisions, and maintain meaningful human oversight.
That last point carries real weight for professional liability. Under the Supreme Court’s policy, lawyers are held accountable for their final work product, whether a person or a program drafted it. A tool chosen without proper safeguards, or output filed without careful review, can lead to client harm, sanctions, or a disciplinary complaint. A documented framework helps show that a lawyer acted reasonably.
For attorney risk management support with AI tools, contact the professional liability team at ISBA Mutual Insurance Company.
