Can an Oregon lawyer use AI and generative AI tools in their legal practice?

Plain-English summary

The opinion answers two questions, both "Yes, qualified": Oregon lawyers may use AI tools, and may use generative AI (GAI) tools, in their legal practice. The qualification is that several Rules of Professional Conduct apply at once.

On competence (Oregon RPC 1.1), the opinion concludes a lawyer may use AI only after taking reasonable steps to understand the benefits and risks of the specific tool, and treats this as an ongoing duty because the tools and their risks change constantly. On communication (RPC 1.2 and 1.4), the opinion concludes that disclosure to the client may be required where the use of AI deviates materially from the client's expectations about the scope of representation, and it lists factors (novelty of the tool, risk to client data, reliance on the tool, effect on fees) that determine whether disclosure is needed.

On fees (RPC 1.5), the opinion concludes lawyers may charge for case-specific research and drafting time using AI but may not duplicate charges or inflate billable hours when AI saves time, and may pass through actual AI costs only after informing the client, preferably in writing. On confidentiality (RPC 1.6), the opinion draws a line between "open" models that learn from inputs and "closed" models that isolate data: using an open model with client information requires the client's informed consent, and without consent the lawyer must anonymize or redact sensitive information. Even closed models require careful review of the vendor's contract terms on data handling, encryption, storage, and destruction.

On supervision (RPC 5.1, 5.2, 5.3), the opinion concludes managerial lawyers must set clear firm policies on AI use and supervisory lawyers must train and oversee subordinate lawyers and nonlawyers, including on the difference between open and closed models. On candor (RPC 3.3, 4.1), the opinion concludes lawyers must verify AI-assisted work and check for court rules restricting or requiring disclosure of AI use.

In practice

Under this opinion, an Oregon lawyer who uses an AI or generative AI tool is acting permissibly only if they have learned enough about that specific tool to understand its risks, have confirmed (or contractually secured) that client information stays confidential, and bill in a way that reflects any time the tool saved rather than duplicating charges. The opinion makes client informed consent a precondition for putting client information into an open, learning model.

The opinion frames the analysis as fact-specific: whether a lawyer must disclose AI use to a client, and whether a particular billing or confidentiality measure is adequate, turns on the factors the opinion identifies rather than a single rule.

Common questions

Q: Can an Oregon lawyer use ChatGPT or other generative AI for client work?

A: Yes, qualified. The opinion concludes generative AI use is permitted if the lawyer is competent in the tool, protects confidentiality, and supervises the output. Putting client information into an open model that learns from inputs requires the client's informed consent.

Q: Does an Oregon lawyer have to tell the client they are using AI?

A: Sometimes. Under RPC 1.2 and 1.4, the opinion concludes disclosure is required when the use of AI deviates materially from the client's reasonable expectations about the scope of representation, judged by factors such as the novelty of the tool and the risk to client data. Routine uses like word processing are generally impliedly authorized.

Q: Can a lawyer bill for time saved by AI?

A: No. The opinion concludes lawyers may bill for actual case-specific time spent but may not duplicate charges or inflate billable hours when AI produces significant time savings, citing ABA Formal Opinion 93-379.

Q: What is the difference between an open and closed AI model under this opinion?

A: An open model continuously learns from data inputs and may incorporate and later disclose sensitive client data; a closed model isolates ingested data in a secure environment. The opinion concludes using an open model with client information requires informed consent, while a closed model still requires careful vendor-contract review.

Background and rules framework

The opinion interprets the Oregon Rules of Professional Conduct as applied to AI tools, drawing on Oregon RPC 1.1 (competence, paralleling Model Rule 1.1 and its Comment 8 on technology), 1.2 and 1.4 (scope and communication), 1.5 (fees), 1.6 (confidentiality), 5.1 through 5.3 (supervision), and 3.3 and 4.1 (candor and truthfulness). It relies heavily on ABA Formal Opinion 512 (2024) on generative AI and ABA Formal Opinion 477R (2017) on securing client communications.

Citations and references

Rules of Professional Conduct:

• Oregon RPC 1.1 / Model Rule 1.1 (competence, including technology)
• Oregon RPC 1.2, 1.4 / Model Rule 1.4 (communication and scope)
• Oregon RPC 1.5 / Model Rule 1.5 (fees)
• Oregon RPC 1.6 / Model Rule 1.6 (confidentiality)
• Oregon RPC 5.1, 5.2, 5.3 / Model Rules 5.1, 5.3 (supervision)
• Oregon RPC 3.3, 4.1 / Model Rules 3.3, 4.1 (candor and truthfulness)

Cases:

• Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), sanctioning lawyers who filed AI hallucinations
• Attorney Grievance Commission of Maryland v. Manger, 396 Md. 134, 913 A.2d 1 (2006), general background research not chargeable to client

Other opinions cited:

• ABA Formal Op. 512 (2024): generative AI tools
• ABA Formal Op. 477R (2017): securing communication of protected client information
• ABA Formal Op. 93-379 (1993): billing for professional fees and expenses
• Florida Bar Ethics Op. 24-1 (2024): generative AI

See also

• ABA Formal Op. 512: Generative AI Tools
• FL Bar Ethics Op. 24-1: Generative AI in Practice
• TX Ethics Op. 705: Generative AI in the Practice of Law
• Alabama Ethics Op. 2026-01: AI Use Best Practices

Source

• Landing page: https://www.osbar.org/ethics/toc.html
• Original PDF: https://www.osbar.org/_docs/ethics/2025-205.pdf