Ethics & Practice
The ethics rules for AI in law practice in North Carolina, and what the NC State Bar expects of family-law firms
If you are weighing AI in law practice in North Carolina, the short answer is that you can use it. The NC State Bar has not banned it. But the NC State Bar's evolving ethics guidance makes clear that your existing duties of confidentiality, competence, and supervision apply to every AI tool you bring into your family-law practice. CapeFear.ai builds AI systems specifically for law firms that meet those obligations by design, so you get the efficiency without the professional risk.
The Ethics Landscape
Where the NC State Bar Stands on AI
The NC State Bar has not yet issued a stand-alone formal ethics opinion dedicated to generative AI, but its 2024 guidance activity leaves little ambiguity about how existing rules apply. The Bar has signaled through committee guidance and national-ethics-opinion tracking that Rules 1.1 (Competence), 1.6 (Confidentiality), 5.1 and 5.3 (Supervision of lawyers and non-lawyers), and 1.4 (Communication) all reach AI-assisted work.
Rule 1.1 now includes a duty to keep current with "changes in the law and its practice, including the benefits and risks associated with relevant technology." That language, borrowed from the ABA's 2012 competence comment, is the basis on which most state bars, including NC, are reading an affirmative obligation to understand the tools you use. Using AI without understanding how it handles client data, how it generates text, or how to verify its output is a competence problem, not just an IT problem.
Rule 1.6 is where family-law practitioners feel the sharpest pressure. Client files in divorce and custody matters contain financial disclosures, mental-health history, allegations of abuse, and minor-child information. Sending any of that to a public, third-party AI service likely constitutes disclosure of confidential information without informed client consent, unless the tool has a binding data agreement that prevents training on your input and limits how the data is retained.
The Bar's stance tracks the consensus forming across state bars nationally. New York, Florida, California, and the ABA's own Formal Opinion 512 (2024) all reach the same practical conclusion: the attorney is responsible for what the AI does, you must disclose AI use to clients in appropriate circumstances, and you cannot outsource your judgment to a system you do not supervise.
The ChatGPT Problem
Why Generic Chatbots Create Real Risk for NC Attorneys
The concern is not AI itself. The concern is where the data goes when you use a consumer-grade tool.
When an attorney pastes a client's financial affidavit or a declaration about a custody dispute into a public chatbot, that text typically leaves the firm's control entirely. Depending on the product's default settings, the conversation may be used to improve the model, stored on servers outside any signed data agreement, or accessible to the provider's staff for safety review. None of that is hypothetical: it is disclosed, in small print, in most consumer terms of service.
For a family-law firm, the consequences of a Rule 1.6 violation are serious. Discipline is one risk. The deeper risk is that a client learns their most sensitive disclosures, the ones they trusted you with, traveled through a third-party system they never agreed to. That is a client-relationship problem before it is ever a Bar complaint.
The fix is not avoiding AI. The fix is using AI inside a controlled environment that does not leak data outward.
No training on your data
A properly structured system processes your client information only for your firm's task and does not use it to train shared models. Your client's financial disclosures stay yours.
A signed data agreement
For practices handling health or mental-health information, a Business Associate Agreement (BAA) is often required. For any confidential client data, a signed data-processing agreement is the minimum that makes Rule 1.6 compliance defensible.
Attorney review before anything goes out
AI drafts a letter, a deadline list, or a document checklist. The attorney sees it, approves it, and it goes. Nothing is sent or filed without a human decision. That is how Rule 5.3 supervision actually works in practice.
A Responsible Approach
What "Ethics-Aligned AI" Actually Looks Like for a Family-Law Firm
CapeFear.ai designs every system for law firms around three operating principles that map directly to the Bar's requirements.
Your tenant, your data. Everything, the agents, the workflow logic, the document templates, the audit logs, lives in an environment provisioned for your firm alone. There is no shared pool where your client data sits alongside another firm's. You can see every action the system has taken, which is exactly the kind of audit trail that demonstrates competent supervision.
Human-in-the-loop by default. Anything client-facing, a letter, a discovery request checklist, a follow-up to opposing counsel, is drafted and queued. The attorney or paralegal reviews it, edits it if needed, and approves it before it moves. The AI accelerates the work; it does not replace the professional judgment. This is not a setting you toggle on. It is how the system is built.
Security infrastructure, not security theater. CapeFear.ai was founded on 14 years of cybersecurity experience. The platform is built toward SOC 2 alignment, with role-based access, encrypted data at rest and in transit, and signed data agreements as a baseline, not an upsell. For matters involving health or mental-health information, a BAA is available.
The practical result is a firm that can use AI for document intake, discovery organization, client intake, deadline tracking, and routine correspondence, while staying inside the rules. The technology does the repetitive work. The attorney stays in charge of every decision that matters.
If you are a North Carolina family-law firm thinking through how to bring AI into your practice without creating a Bar complaint or a client-trust problem, the conversation starts with the systems question, not the AI question. We are happy to walk through what that looks like for your firm in a 30-minute call, no pressure, no sales pitch. Visit our AI for law firms page to learn more or to schedule time.
Before You Sign
A short vendor-vetting checklist for AI in a North Carolina law firm
The NC State Bar's competence and supervision duties (Rules 1.1 and 5.3) mean the responsibility for an AI vendor sits with the attorney, not the vendor. Rule 5.3 governs your supervision of nonlawyer assistance, and an AI tool is nonlawyer assistance under that rule. Before you put any client information through an AI tool, work through these questions. If a vendor cannot answer them in writing, that is your answer.
- Will my data train shared models? The answer you want is no. Your client's financial affidavits and custody declarations should be processed only for your firm's task, never folded into a model other firms use.
- Will you sign a data-processing agreement, and a BAA if I need one? For confidential client data, a signed data-processing agreement is the floor for a defensible Rule 1.6 position. For matters touching mental-health or medical records, a Business Associate Agreement is often required.
- Where does the data live, and who can see it? Ask where data is stored, whether it leaves the country, and whether vendor staff can read it. Map the answers against your Rule 1.6 confidentiality duty.
- Does an attorney review output before anything goes out? Rule 5.3 supervision means a human approves AI-drafted letters, checklists, and filings before they move. A tool that sends or files on its own does not fit that rule.
- Is there an audit log? A record of every action the AI took is how you demonstrate competent supervision if a client, opposing counsel, or the Bar ever asks.
- Do I own what gets built, and can I leave? Confirm the agents, prompts, and workflows live in an environment you control, with no lock-in that traps your client data inside a vendor's platform.
Common questions
Questions firms ask before they start.
Is the NC State Bar okay with attorneys using AI?
Yes, with conditions. The NC State Bar has not banned AI, but its guidance makes clear that the existing rules on competence (Rule 1.1), confidentiality (Rule 1.6), and supervision (Rules 5.1 and 5.3) all apply to AI-assisted work. Attorneys must understand the tools they use, protect client data, and supervise AI output before it goes to clients or courts.
Can I use ChatGPT for client work as an NC attorney?
Not without serious caution. Consumer AI tools like ChatGPT do not come with a signed confidentiality or data-processing agreement, and your input may be stored or used in ways you cannot control. Pasting client information into a public chatbot very likely violates Rule 1.6 unless you have obtained informed client consent for that disclosure. Most NC attorneys should avoid consumer-grade AI for anything containing client-specific information.
What is a BAA and do I need one for AI in my law firm?
A Business Associate Agreement (BAA) is a contract required under HIPAA when you share protected health information (PHI) with a vendor. Family-law matters that involve mental-health records, substance-abuse treatment, or medical evaluations may touch PHI. If they do, any AI vendor processing that information needs a signed BAA. Even outside HIPAA, a signed data-processing agreement that limits how the vendor uses your client data is the minimum standard for responsible AI use.
Do I have to tell clients I am using AI?
The Bar has not issued a bright-line rule, but the trend in ethics guidance nationally, and consistent with NC Rule 1.4 (Communication), is that attorneys should disclose AI use when it is material to the representation or when the client would reasonably want to know. Using AI to draft a letter the client signs, or to organize their documents, are the kinds of uses where disclosure is becoming standard practice.
How does CapeFear.ai keep my client data confidential?
Every law firm we work with gets its own isolated environment. Client data is never shared with other firms or used to train shared AI models. All data is encrypted at rest and in transit. We sign a data-processing agreement as a baseline, and a BAA is available for matters involving protected health information. Every AI action is logged so attorneys can demonstrate supervision if it is ever questioned.
What tasks can AI actually help a family-law firm with?
Practical, low-risk starting points include organizing discovery documents and tracking what has and has not been produced, drafting routine client correspondence for attorney review, building and maintaining deadline checklists from scheduling orders, and handling initial intake questions from prospective clients. In each case, the attorney reviews and approves the output before it goes anywhere.
Substitution vs. augmentation
Why "the attorney decides" is the whole point.
The competence and supervision duties the Bar cares about come down to a single design question: does the tool let you skip your own judgment, or does it protect it? The harm from AI tracks effort substitution, letting the tool do the task instead of you, not AI assistance itself. Use it to gather and organize faster and your skill is intact. Hand it the judgment and that's the judgment you stop sharpening. So we built the service so it can't substitute for yours: it drafts and proposes, and a person reviews, edits, and approves before anything moves.
That effort-substitution effect isn't unique to law. A 2026 study of roughly 26,800 students over about 30 months found AI use cut later exam scores by around 20%, with the damage concentrated among the roughly 80% who let the AI do the task rather than speed up their own work. It's one supporting data point, not proof about legal work, and "students aren't attorneys" misses the point: effort substitution dulls any skill, which is exactly why human sign-off is built in rather than left as a setting.