AI for Law Firms in 2026: A Portland Guide to Using AI Without Risking Client Confidentiality

Somewhere in your firm, someone has already pasted a paragraph from a client email into a free chatbot to “make this sound more professional.” Maybe it was a paralegal tightening up a demand letter, or an associate summarizing a 60-page deposition transcript at 11 p.m. That is what AI for law firms looks like in 2026 for most small practices: not a robot lawyer, but everyday drafting and summarizing tools quietly handling privileged information in accounts the firm does not control.

For solo practitioners and small firms in Portland, Lake Oswego, Beaverton, Hillsboro, Clackamas, and Vancouver WA, generative AI can genuinely save hours on first drafts, summaries, and routine correspondence. It also introduces two risks that are specific to legal work: exposing confidential client information, and filing work product that contains made-up case law. Both are avoidable with clear rules and a few IT controls.

This post is the legal-industry entry in our AI series. For general tool choices, see AI productivity tools and policies for small business. For the threat side, see AI cybersecurity for small business. If you are about to turn on Copilot, read Microsoft 365 Copilot security first. And if you need a written rulebook, start with our AI acceptable use policy guide. This article stays focused on what is different for law firms: privilege, ethics duties, trust accounts, and court filings.

What AI for law firms should mean in 2026

For a small or mid-size practice, useful AI for law firms usually falls into a handful of categories:

• First drafts of routine letters, engagement letter sections, client updates, and internal memos
• Summaries of long documents, email threads, transcripts, and discovery sets you are allowed to process
• Rewriting and plain-language translation of dense legal text for clients
• Issue spotting and outlines that an attorney then researches and verifies
• Practice management help, such as intake form cleanup, calendar language, and marketing copy

What it should not mean: letting a general-purpose chatbot do legal research you file without checking, pasting client files into consumer AI accounts, or letting AI send anything to a client, court, or opposing counsel without an attorney reading it first.

The rule of thumb we give every firm is simple. AI drafts. Attorneys decide. The lawyer still owns the advice, the filing, the citation, and the signature.

The ethics rules already apply to AI

You do not need to wait for a new AI-specific rule to know what is expected. In July 2024, the American Bar Association issued Formal Opinion 512 on generative AI tools. It does not create new obligations; it explains how the existing Model Rules apply when lawyers use AI. The main themes are ones every managing partner will recognize:

• Competence (Rule 1.1): Lawyers need a reasonable understanding of what the AI tool does, what it is bad at, and how it handles data. You do not need to be an engineer, but “I didn’t know it made things up” is not a defense.
• Confidentiality (Rule 1.6): Before entering client information into an AI tool, lawyers need to understand whether that information could be disclosed, retained, or used by the provider. The opinion indicates that informed client consent may be needed before using self-learning tools with client information.
• Communication (Rule 1.4): In some situations, clients should be told how AI is being used on their matter.
• Candor and meritorious claims (Rules 3.1, 3.3): Anything filed with a court has to be accurate, which means every AI-assisted citation and quote must be verified.
• Supervision (Rules 5.1 and 5.3): Partners and supervising lawyers need policies and training so associates, paralegals, and staff use AI properly.
• Fees (Rule 1.5): Billing has to be reasonable. Charging for hours AI saved, or billing clients for time spent learning a general tool, raises questions.

Oregon and Washington each have their own Rules of Professional Conduct based on the ABA Model Rules, and both state bars publish ethics guidance. The ABA opinion is not binding on its own, so check the Oregon State Bar and Washington State Bar Association resources for your jurisdiction, or talk to your ethics counsel. Our role at PDX Information Technology Services is the technical side: making sure the tools, accounts, and data flows support the duties your attorneys already have.

Risk 1: Client confidentiality and where your prompts go

The biggest practical risk is not a dramatic hack. It is a well-meaning staff member pasting privileged material into a tool the firm never approved.

Consumer AI vs business AI accounts

Not all AI accounts handle data the same way. Free and personal consumer accounts are generally governed by consumer terms, which can allow the provider to retain conversations and, depending on settings, use them to improve their models. Business and enterprise plans from the major providers typically offer stronger commitments, such as not training on your data, admin controls, and defined retention.

The details change often, so do not rely on a blog post (including this one) for the current terms. What matters for your firm:

1. Decide which AI tools are approved, under firm-owned business accounts.
2. Read the data use, retention, and training terms for those specific plans, and keep a copy with your vendor files.
3. Block or discourage personal accounts for firm work, especially on firm devices.

The “never paste” list for law firms

Even in approved tools, some information should stay out unless the firm has made a deliberate decision and, where needed, gotten client consent. A starting list:

• Client names combined with matter facts, especially in sensitive practice areas (family law, criminal defense, immigration, employment, estate planning)
• Settlement figures, negotiation strategy, and litigation strategy memos
• Social Security numbers, financial account numbers, medical records, and anything under a protective order
• Sealed or confidential court documents
• Trust account details and wire instructions
• Opposing party data you obtained in discovery under restrictions

A good habit is anonymizing before prompting: “Client A, a small manufacturer in Washington County” instead of real names and identifying details. It is not perfect, but it removes the most dangerous part of most prompts.

Shadow AI on phones and home computers

Small firms often have attorneys who work from home, from court, and from their phones. That is where shadow AI grows: a personal ChatGPT app on a personal phone, a browser extension that “summarizes any page,” a free transcription app recording a client call. None of it shows up in your firm’s records, and none of it is covered by your vendor agreements.

The fix is a mix of policy and tooling: an approved tool list, managed devices, browser extension controls, and Microsoft 365 policies that make it harder for firm data to leave firm-managed apps. Our AI acceptable use policy guide has a template you can adapt for attorneys and staff.

Risk 2: Hallucinated citations and AI-generated errors

General-purpose chatbots can produce case names, quotes, and citations that look completely real and do not exist. This is not hypothetical. In the widely reported 2023 federal case Mata v. Avianca in New York, attorneys were sanctioned after filing a brief containing fake cases generated by ChatGPT, and courts have continued to address similar filings since then.

For a small firm, the safeguards are procedural more than technical:

• Verify every citation in a trusted legal research platform before it goes into any filing or client advice.
• Never cite a case you have not read. If AI suggested it, pull it and read it.
• Check court rules and standing orders. Some judges require disclosure or certification of AI use in filings. Your docketing and filing checklist should include a step for this.
• Label AI-assisted drafts internally so the reviewing attorney knows to check more carefully.
• Prefer legal-specific research tools for research tasks, and use general chatbots for writing and summarizing, not for finding law.

Hallucinations also show up outside court filings: a wrong deadline in a client update, a misquoted contract term in a summary, or an invented statute section in a blog post. Every AI output that leaves the firm needs a human check.

Risk 3: AI-powered fraud aimed at law firms

Law firms are attractive targets because they move money, especially in real estate, settlements, and estate work, and because they communicate constantly by email with people they have never met in person. AI makes the scams more convincing.

Wire fraud and trust account scams

Business email compromise (BEC) against law firms usually looks like a client, title company, or opposing counsel “updating” wire instructions at the last minute. AI writing tools have removed the bad grammar that used to give these emails away. The message can now match the tone of a real sender almost perfectly.

Controls that work:

• Call-back verification for any change to wire instructions, using a phone number you already have on file, never one from the email
• Two-person approval for outgoing wires from trust or operating accounts
• Advanced email security that looks at sender behavior, lookalike domains, and impersonation. We use INKY for this with our clients, layered on top of Microsoft 365’s built-in protections.
• Multi-factor authentication on every mailbox, so a stolen password alone does not hand over your email

Deepfake voice and video

Voice cloning means a “client” or “managing partner” can call and sound right. Treat urgent, secret, or unusual payment requests the same way regardless of how they arrive: verify through a second, known channel before acting. A short code word for internal payment approvals is a cheap, effective control for small firms.

AI-written phishing to steal credentials

Phishing emails that mimic court notices, e-filing systems, document-sharing links, and bar association messages are common. Security awareness training should include examples aimed at legal staff, and email security should check links at the time they are clicked, not just when the email arrives. Our AI cybersecurity guide covers these attack patterns in more detail.

Copilot and Microsoft 365: fix permissions before you turn it on

Many Portland-area firms run on Microsoft 365, and Microsoft 365 Copilot is an appealing option because it works inside Outlook, Word, Teams, and SharePoint. Microsoft states that Copilot respects existing user permissions and that prompts and responses for Microsoft 365 Copilot are not used to train its foundation models. That is helpful, but it also means Copilot will surface anything a user can technically access.

For a law firm, that is the core issue. If your SharePoint or shared drives are set up so everyone can see everything, Copilot can summarize a sensitive matter for someone who was never supposed to be on it, including matters behind an ethical wall.

Before a Copilot rollout, a small firm should:

1. Audit SharePoint, OneDrive, and Teams permissions. Find “Everyone” and “anyone with the link” sharing on matter files.
2. Organize matters by access. Matter-based sites or libraries with membership limited to the assigned team make ethical walls enforceable.
3. Apply sensitivity labels (Microsoft Purview) to privileged or highly confidential documents where your licensing supports it.
4. Set data loss prevention rules for obvious sensitive data such as Social Security and financial account numbers.
5. Start with a pilot group of a few attorneys and staff, then expand.
6. Review licensing. Copilot add-ons and security features depend on your Microsoft 365 plan.

The detailed checklist is in our Microsoft 365 Copilot security post.

Vetting legal AI vendors

Legal-specific AI products (for research, contract review, e-discovery, and practice management) are multiplying. Before you sign, ask each vendor:

• Is our data used to train your models? Can we get that commitment in the contract?
• Where is data stored, how long is it retained, and how is it deleted when we leave?
• Who at the vendor can access our data, and is access logged?
• Do you support single sign-on and multi-factor authentication through our Microsoft 365 accounts?
• Do you have an independent security assessment such as a SOC 2 report?
• What happens to our data if you are acquired or shut down?
• Can we control which users and matters the tool can see?

Keep the answers with your vendor records. If a client or the bar ever asks how you protected their information, documented due diligence is much better than “it seemed fine.”

A 30-day AI plan for a small law firm

Here is a practical starting plan for a firm with roughly 3 to 30 people.

Week 1: Find out what is already happening

• Ask attorneys and staff, without blame, which AI tools they already use and for what
• Review sign-ins and browser extensions on firm devices
• List every place client data lives: email, document management, practice management, SharePoint, local drives

Week 2: Decide the rules

• Pick approved tools under firm-owned business accounts
• Write a short AI acceptable use policy covering the never-paste list, citation verification, client communication, and who to ask
• Decide how your engagement letters will address AI use, with input from ethics counsel

Week 3: Put IT guardrails in place

• Enforce multi-factor authentication everywhere
• Lock down SharePoint and OneDrive sharing, and set up matter-based access
• Deploy email security and wire verification procedures
• Manage devices so firm data stays in firm-managed apps

Week 4: Train and pilot

• Run a short training for attorneys and staff with real examples, including a fake citation exercise
• Pilot approved AI tools with a small group
• Schedule a review in 90 days to adjust the policy and tool list

How PDX IT helps Portland-area law firms with AI

PDX Information Technology Services supports small and mid-size businesses across the Portland metro, including law firms that need tight security without a full-time IT department. For firms adopting AI, that typically includes:

• Microsoft 365 security reviews and permission cleanup before Copilot
• Email security with INKY, multi-factor authentication, and phishing training
• Device management for attorneys working from the office, home, and court
• Help evaluating AI vendors from a security and data-handling perspective
• Backup and recovery for document and practice management data

You can learn more about our managed IT for law firms in Portland, our broader managed IT services, and our compliance services. On-site support is available from Hillsboro to Gresham, down to Wilsonville, and across the river to Vancouver WA, with no trip fee in our service area.

If you are comparing providers, our guide on what to look for in managed IT services in Portland and our breakdown of managed IT services cost in Portland are good next reads.

Frequently asked questions

Can lawyers use ChatGPT for client work?

Lawyers can use generative AI tools, but the ethics rules on competence, confidentiality, supervision, and candor still apply. ABA Formal Opinion 512 explains how those duties apply to AI. In practice, that means using firm-approved business accounts, understanding how the tool handles data, keeping confidential information out unless the firm has decided it is appropriate (and obtained client consent where needed), and verifying every output. Check Oregon or Washington bar guidance for your jurisdiction.

Is Microsoft 365 Copilot safe for law firms?

Copilot can be a reasonable choice because it works within your Microsoft 365 tenant and respects existing permissions. The risk is that it respects whatever permissions you have today, including overly broad ones. Clean up SharePoint, OneDrive, and Teams access, set up matter-based permissions, and pilot it with a small group before a firm-wide rollout.

How do we stop staff from using unapproved AI tools?

Combine a clear written policy with technical controls. Provide approved tools so people have a good option, manage firm devices, control browser extensions, and use Microsoft 365 policies to limit copying firm data into unmanaged apps. Training matters too: most shadow AI use comes from people trying to work faster, not from bad intent.

Do we need to tell clients we use AI?

It depends on how you use it and on your jurisdiction’s guidance. ABA Formal Opinion 512 discusses situations where client communication or informed consent may be required, such as when client information is entered into certain AI tools. Many firms are adding AI language to engagement letters. Talk to your ethics counsel about the right approach for your practice.

What is the biggest AI security risk for a small law firm?

For most small firms, the two biggest risks are confidential client data going into unapproved AI accounts and AI-assisted wire fraud targeting trust and real estate transactions. Both are addressed with clear policies, firm-owned AI accounts, multi-factor authentication, strong email security, and call-back verification for any payment change.

Can AI-generated legal research be trusted?

Not without verification. General-purpose chatbots can invent cases and quotes that look real. Every citation should be pulled and read in a trusted legal research platform before it is used in a filing or in client advice, and attorneys should check court rules for any AI disclosure requirements.

Talk to a local IT team about AI at your firm

If your firm is trying to figure out where AI fits, or you suspect it is already being used without guardrails, we are happy to talk it through. There is no pressure and no long pitch, just a practical conversation about your setup.

• Book a time with Steve on Calendly
• Email sales@pdxittech.com
• Call 971-331-4871

This article is general information about IT and security practices. It is not legal advice. For ethics questions, consult your state bar’s guidance or your own ethics counsel.