AI Governance for Law Firms: Ethics, Risk and Policy
Published · Last updated · 8 min read

AI governance in a law firm is the set of policies, approvals, training and technical controls that decide which AI tools lawyers and staff may use, with what client information, and how output is checked before it reaches a client or court. ABA Formal Opinion 512 applies existing duties of competence, confidentiality, communication, supervision, candor and reasonable fees to generative AI, so governance is an ethics requirement, not an option.
Key takeaways
- ABA Formal Opinion 512 (July 29, 2024) creates no new rules; it applies Model Rules 1.1, 1.4, 1.5, 1.6, 3.1, 3.3, 5.1 and 5.3 to generative AI.
- Informed client consent is required before entering client information into a self-learning AI tool that could disclose it.
- Every AI-assisted filing needs independent verification. On February 18, 2026, the Fifth Circuit sanctioned a lawyer $2,500 over a brief with 21 fabricated quotations or serious misrepresentations.
- Managing lawyers are expected to set clear policies on permissible AI use and to supervise lawyers, staff and vendors under Rules 5.1 and 5.3.
- Good governance covers an approved tool list, data rules, verification steps, training, vendor review and logging.
Why does AI governance matter for law firms now?
Lawyers are using generative AI for research, drafting, summarizing discovery, transcription and client communications, often through tools adopted by individuals rather than the firm. Without governance, client information can end up in consumer tools with unclear retention terms, and unverified output can end up in filings.
Courts have run out of patience with the second problem. In Fletcher v. Experian Information Solutions, decided February 18, 2026, the Fifth Circuit found that counsel used AI to draft a substantial portion, if not all, of a reply brief, identified 16 fabricated quotations and 5 serious misrepresentations of law or fact, and sanctioned her $2,500. The court noted it would likely have imposed a lesser sanction had she been candid about the AI use when asked, and said that pleading ignorance of the risks of unverified AI drafting is no longer an excuse.
What does ABA Formal Opinion 512 require?
Formal Opinion 512, the ABA's first formal opinion on generative AI, applies existing duties to these tools:
- Competence (Rule 1.1). Lawyers need not be AI experts, but must reasonably understand the capabilities and limitations of the specific tool they use and review output with a degree of independent verification appropriate to the task.
- Confidentiality (Rules 1.6, 1.8(b), 1.9(c) and 1.18(b)). Lawyers must assess the risk that client information entered into a tool could be disclosed or accessed, inside or outside the firm. For self-learning tools that could reveal client information, informed consent is required first, and boilerplate in an engagement letter is not enough.
- Communication (Rule 1.4). Disclose AI use when a client asks, when the engagement terms or outside counsel guidelines require it, or when AI use is relevant to the client's decisions about the representation.
- Meritorious claims and candor (Rules 3.1 and 3.3). Lawyers must review AI output for accuracy before relying on it in filings and correct errors.
- Supervision (Rules 5.1 and 5.3). Managerial lawyers should establish clear policies on permissible AI use and make sure lawyers, staff and vendors follow them.
- Fees (Rule 1.5). Lawyers billing hourly may bill only the time actually spent, and generally may not bill clients for the time to learn a general-purpose tool.
What should a law firm AI policy include?
- Approved tools. A short list of tools reviewed by the firm, with consumer and personal accounts prohibited for client work.
- Data rules. Which categories of information may be used with which tools, and when client consent is required.
- Verification. Every citation, quotation and factual statement produced with AI is checked against the source before it leaves the firm, and the reviewing lawyer is identified.
- Disclosure. When and how AI use is disclosed to clients and courts, including any court standing orders or local rules.
- Billing. How AI-assisted work is recorded and billed consistent with Rule 1.5.
- Training. Required training before access, refreshed as tools change.
- Incident handling. What to do if client information is entered into an unapproved tool or an error reaches a filing, including prompt correction with the court.
How should firms evaluate AI tools and vendors?
Formal Opinion 512 expects the same due diligence lawyers apply to any vendor handling client information. Before approving a tool, ask:
- Does the provider use your inputs or outputs to train models available to others, and can that be turned off contractually?
- How long are prompts, files and outputs retained, where are they stored, and who at the provider can access them?
- Does the tool support single sign-on, MFA, role-based access and audit logs?
- Can matter-level permissions and ethical walls be enforced inside the tool?
- What do the contract terms say about confidentiality, security incidents, notice and liability?
- Has the firm tested the tool on its own work to understand accuracy and limits?
The NIST AI Risk Management Framework offers a voluntary structure for mapping, measuring and managing these risks if your firm wants a formal reference.
What cybersecurity risks does AI add for law firms?
- Data leakage. Client information pasted into unapproved tools, browser extensions or meeting assistants.
- Over-broad access. AI assistants connected to email and document systems can surface files a user should not see if permissions are loose.
- More convincing phishing. AI-written messages and cloned voices make wire fraud and impersonation harder to spot. Training and payment verification procedures need updating.
- Prompt injection. Documents or web pages can contain hidden instructions that manipulate AI tools connected to firm data.
- Shadow AI. Tools adopted by individuals without review, outside the firm's logging and retention controls.
Most of these are addressed by the same controls that protect the firm generally: identity management with MFA, least-privilege access, data loss prevention, logging and staff training. Our guide to cybersecurity for law firms covers that foundation.
How do state bars and courts add to the ABA guidance?
Many state bars have issued their own AI ethics opinions or guidance since 2024, generally tracking Formal Opinion 512 with local variations, and some states are amending the comments to their professional conduct rules. Individual courts and judges have also issued standing orders on AI use in filings. These requirements change often, so assign someone to track the jurisdictions and courts where your lawyers practice and update the policy when they change. When a court order and the firm policy differ, the stricter rule should control for that matter.
What are the steps to build AI governance in a firm?
- Find out what is already in use. Survey lawyers and staff and review network and account data for AI tools.
- Form a small committee. A partner, the firm administrator, IT and someone responsible for risk and ethics.
- Approve a starting set of tools after vendor review, and block or retire the rest.
- Adopt the written policy and update engagement letters and consent processes where needed.
- Configure controls: single sign-on, MFA, permissions, retention and logging.
- Train before granting access, with examples of verification failures like the Fifth Circuit case.
- Review quarterly. New tools, new rules and lessons from near misses.
How can NorthStar help with AI governance?
NorthStar helps law firms adopt AI safely: assessing current use, reviewing tools and vendors, configuring identity, access and data controls, and training lawyers and staff. See our AI advisory services and law firm IT services. We support the technical and operational side; ethics and legal judgments remain with the firm.
Start with our free security check, read about ABA ethics and cybersecurity, or browse more legal industry resources.
What else do law firms ask about AI governance?
Do we have to tell clients we use AI?
Not in every case. Under Formal Opinion 512, disclosure is required when a client asks, when engagement terms or outside counsel guidelines require it, or when AI use matters to the client's decisions. Informed consent is required before entering client information into self-learning tools that could disclose it.
Can we bill clients for AI-assisted work?
Yes, consistent with Rule 1.5. Hourly billing must reflect time actually spent, and firms generally may not charge clients for time spent learning a general-purpose AI tool. Flat fees must remain reasonable.
Is a general engagement letter clause enough for AI consent?
Formal Opinion 512 says boilerplate language in an engagement letter is not sufficient for informed consent. The client needs enough explanation of the risks and benefits to make an informed decision.
Who is responsible if a junior lawyer files an AI-generated brief with errors?
The signing lawyer is responsible for the filing, and supervising lawyers can face responsibility under Rule 5.1 if the firm lacked reasonable policies and supervision.
Sources
- ABA Formal Opinion 512: Generative Artificial Intelligence Tools (July 29, 2024)
- Fletcher v. Experian Information Solutions, No. 25-20086 (5th Cir. Feb. 18, 2026)
- ABA Model Rule 1.6: Confidentiality of Information
- NIST AI Risk Management Framework
Industry Resources
AI Advisory for Law Firms
NorthStar helps law firms adopt AI with clear policies, reviewed tools and the security controls client confidentiality requires.
Learn More →About the author

Ken Satkunam, CISM
President & Founder, NorthStar Technology Group
Ken has spent over 25 years in IT leadership, serving in roles from technical support to CIO for organizations as large as 23,000 employees. He founded NorthStar Technology Group in 2000 to help regulated organizations build secure, compliant, and operationally resilient technology environments. Ken holds the Certified Information Security Manager (CISM) credential from ISACA and is the co-author of the Amazon best-seller "Cyber Attack Prevention." He has been quoted in industry publications including eWeek and DM News, and NorthStar has been recognized on the Inc. 5000 list in both 2024 and 2025.