Artificial Intelligence (AI) in Legal Practice: Ethics, Risk, and Compliance Guidelines for Attorneys

Types : Alerts

By: Ross Fox, Esq. of Montgomery McCracken, Dylan Henry, Esq. of Montgomery McCracken, and John Puszcz, Esq. of CP Engineers

1. The AI Revolution in Law: Are You Ready?

AI has moved from emerging technology to an everyday tool in the legal practice. Attorneys across all practice areas now routinely use generative AI to summarize documents, conduct legal research, draft memoranda, review contracts, and streamline discovery. As these tools mature, their role will only expand.

Courts, bar associations, and regulators recognize that AI improves efficiency and reduces costs. But they have delivered a clear message: AI is a tool to assist attorneys, not a replacement for professional judgment or ethical responsibilities, and attorneys should use AI at their own risk.

With weekly “AI-Use By Lawyer Gone Wrong” headlines, no legal professional can claim ignorance to these risks. The dangers are amplified when AI is used without human oversight, independent verification, and proper safeguards.

This article examines developments in Pennsylvania, New Jersey, New York, and Delaware to illustrate broader national trends. Attorneys in other jurisdictions should review their own rules, ethics opinions, and standing orders.

2. How AI Is Transforming Legal Work

AI is integrated into virtually every phase of legal practice, from research and drafting to discovery, due diligence, contract review, and legal operations.

Litigation: Discovery often requires reviewing thousands or millions of documents. AI-assisted review tools rapidly organize ESI, identify responsive documents, extract key information, and generate summaries that help attorneys understand complex factual records efficiently.

Transactional practice: Attorneys use AI to compare contracts, identify inconsistent provisions, summarize diligence materials, and prepare preliminary drafts.

In-house legal departments: Legal teams are incorporating AI into knowledge management, compliance monitoring, and internal workflows.

AI’s greatest benefit is not replacing attorneys. It frees them to focus on work requiring uniquely human judgment: litigation strategy, client counseling, negotiations, witness preparation, and legal analysis. AI excels at organizing, summarizing, and drafting.

3. Key AI Risks Every Lawyer Must Understand

Despite its benefits, AI presents significant risks that attorneys and legal departments must actively manage.

Hallucinations. AI can fabricate case citations, quotations, factual assertions, and legal analysis that appear entirely plausible despite having no basis in fact. These inaccuracies are often difficult to detect without independent verification (e.g., paraphrasing case holdings), making them particularly dangerous in legal work product.

Overreliance. AI frequently produces polished, persuasive writing that creates unwarranted confidence in its accuracy. The danger is not just that AI makes mistakes, but that those mistakes often appear professionally written and legally sophisticated. As recent sanctions demonstrate, persuasive prose is not a substitute for accurate analysis.

Confidentiality breaches. Attorneys work with privileged communications, proprietary information, trade secrets, and sensitive personal data. Unlike hallucinations, confidentiality failures may expose clients to privilege waivers, cybersecurity risks, regulatory violations, and reputational harm.

Shadow AI. Unauthorized use of consumer AI platforms without organizational oversight presents a fundamentally different challenge. Legal departments may be unaware that confidential information, trade secrets, litigation strategy, or regulated data is being submitted to third-party systems. Even worse, business partners may bypass in-house legal resources entirely and use AI for their own legal analysis. Governance must extend beyond attorneys to encompass enterprise-wide AI usage.

Bias. AI systems may perpetuate implicit bias if trained on datasets reflecting historical or societal biases. AI-generated recommendations should always be evaluated critically, particularly in employment decisions, compliance reviews, or predictive legal analysis.

4. Courts Have Spoken: AI Does Not Replace Attorney Judgment

Courts acknowledge AI’s value as a litigation tool but are clear that its use doesn’t diminish professional obligations. Attorneys remain personally responsible for every AI-generated factual assertion, legal citation, and argument submitted under their signature.

Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), illustrates this approach. Attorneys submitted a brief containing multiple fictitious judicial opinions generated through AI-assisted research. The court imposed sanctions, emphasizing that Rule 11 obligations apply equally to AI-assisted filings and that attorneys remain personally responsible for accuracy.

Mata was not isolated. Courts have imposed meaningful consequences for AI misuse, including:

• Monetary sanctions
• Waiver of attorney-client privilege
• Mandatory continuing legal education
• Waiver of client’s substantive rights

The consequences extend beyond attorney discipline. AI-generated inaccuracies can prejudice client interests by undermining credibility, increasing costs, delaying proceedings, or affecting case outcomes.

Three recurring principles emerge:

1. Human oversight remains essential. AI may assist with researching, drafting, and organizing, but it cannot replace independent legal judgment.

2. Verification is not delegable. Attorneys must confirm what they submit to the court is accurate (e.g., legal authorities exist, accurately state the law, and support the propositions cited).

3. Courts expect accountability, not avoidance. The consensus is not that attorneys should avoid AI. If attorneys choose to use AI, they must do so responsibly and remain accountable.

5. Existing Ethics Rules Already Govern AI Use

Courts and bar associations have generally concluded that existing rules of professional conduct govern AI use.

A. Competence

The duty of competence increasingly encompasses technological competence. Attorneys need not become software engineers, but they should understand the capabilities and limitations of the AI tools they use. This includes knowing when AI should not be used and what safeguards are necessary.

B. Confidentiality

Many AI platforms process prompts on third-party servers, retain user inputs, or use submitted information to train future models. Attorneys should understand how any AI platform stores, processes, and protects information before uploading privileged communications, trade secrets, PII, or other confidential materials.

C. Candor and Communication

The duty of candor remains unchanged. While no statewide obligations require AI disclosure, practitioners must verify individual court rules and standing orders. Certain matters warrant client discussion regarding AI use, particularly where AI materially affects representation or involves sensitive information.

D. Supervision

AI use implicates supervisory responsibilities. Partners, supervising attorneys, and legal departments remain responsible for ensuring that subordinates use AI appropriately.

E. Reasonable Fees

AI raises important billing questions. While AI can reduce time on certain tasks, it does not alter the obligation to charge reasonable fees. AI-generated work still requires attorney review, verification, and refinement. Fees should accurately reflect services provided and remain reasonable under applicable rules.

6. State-by-State AI Guidance: PA, NJ, NY, and DE

Developments in Pennsylvania, New Jersey, New York, and Delaware reveal consistent themes across jurisdictions.

Pennsylvania: No statewide AI disclosure rule exists, but a Joint Formal Opinion from the Pennsylvania and Philadelphia Bar Associations provides a general ethical framework. Requirements vary by county and court. Several Pennsylvania courts have sanctioned attorneys for AI-generated inaccuracies.

New Jersey: No blanket disclosure requirement. The Supreme Court’s preliminary guidance emphasizes that existing Rules of Professional Conduct govern AI use. New Jersey has adopted technology-related CLE requirements reflecting the importance of technological competence.

New York: The most structured approach. Part 161 of the Rules of the Chief Administrator permits AI in court submissions but requires independent review to ensure no hallucinated authorities. Individual judges may impose additional requirements through standing orders.

Delaware: No comprehensive statewide rules, but courts emphasize AI remains a tool, not a substitute for professional judgment. Guidance focuses on competence, confidentiality, and independent verification.

7. Practical AI Guidance for Lawyers and Legal Departments

For all attorneys: Independently verify every AI-generated citation, quotation, factual assertion, and legal proposition before filing. Review applicable local rules and standing orders.

For law firms: Develop written AI policies addressing approved platforms, confidentiality safeguards, supervision, and training. Distinguish between permissible AI uses and tasks requiring heightened attorney involvement.

For in-house legal departments: Address governance beyond attorney use. Consider how AI is used throughout the organization, including by non-legal business partners and outside counsel. Coordinate AI policies among legal, IT, and compliance teams.

8. The Path Forward: Embracing AI Responsibly

AI will remain integral to modern legal practice. Used properly, it improves efficiency and enables attorneys to focus on strategic work. Used improperly, it presents substantial litigation, ethical, and reputational risks. Courts have embraced responsible AI use while making clear that attorneys remain accountable.

As courts refine AI guidance, attorneys should expect increasing emphasis on human oversight, independent verification, and technological competence. Lawyers who understand AI’s capabilities and limitations will be best positioned to leverage its benefits while meeting professional obligations.

The question is no longer whether AI will become part of legal practice. That transformation is underway. The question is whether lawyers, firms, and legal departments will implement the governance necessary to use AI responsibly, ethically, and effectively.

Questions about AI governance for your firm or legal department? Contact Ross Fox or Dylan Henry to discuss strategies for responsible AI adoption.