Artificial intelligence has officially entered the legal profession.

Clients now walk into law offices carrying AI-generated trusts, chatbot-created LLC agreements, automated tax ideas, AI-written legal memoranda, and entire “estate plans” assembled after a weekend of prompting.

At the same time, attorneys themselves are increasingly using artificial intelligence for drafting, legal research, intake summaries, document review, deposition analysis, client communication, marketing content, and practice management.

None of that is automatically bad.

In fact, when used properly, AI can be extraordinarily helpful.

It can help lawyers organize information, summarize transcripts, draft first-pass content, identify issues, create checklists, and reduce administrative friction.

But there is a deeper problem.

AI can produce words.

AI can produce structure.

AI can produce impressive-looking documents.

But AI does not replace judgment.

And in legal planning—especially Estate Planning, Asset Protection, Business Succession, Tax Planning, Trust Administration, and family wealth strategy—judgment is the entire game.

Estate planning is not document generation.

It is control architecture under pressure.

That distinction matters.

---

The Weekend AI Spiral: A Story That Says More Than Statistics

Recently, I spoke with someone who had spent nearly an entire weekend trying to solve their estate plan with artificial intelligence.

At first, it sounded productive.

They had used AI to research trusts, LLCs, tax ideas, offshore planning, asset protection, and even dynasty trust concepts.

But as the conversation continued, something became obvious.

The person was not becoming more certain.

They were becoming more overwhelmed.

Every AI answer generated five more questions.

Every question generated ten more scenarios.

Every scenario created another rabbit hole.

At one point, the individual laughed and said:

“I think I may need AI rehab.”

Funny.

But there was truth in it.

The pattern felt less like legal planning and more like gaming, gambling, or social media scrolling.

One prompt becomes ten.

Ten prompts become three hours.

Three hours becomes emotional attachment to the machine's answer.

Eventually, society may need:

AIA — Artificial Intelligence Anonymous

“Hello, my name is Tom.”

“And I asked AI 6,400 questions about offshore trusts this month.”

The room responds:

“Hi Tom.”

Humor aside, this is becoming a real issue.

Information overload is not the same thing as understanding.

And AI-generated confidence is not the same thing as legal accuracy.

---

AI Sounds Smart Even When It Is Wrong

One of the most dangerous characteristics of artificial intelligence is confidence.

Large language models do not think like lawyers.

They predict language patterns.

That means AI can sound persuasive, structured, and authoritative while still being wrong.

AI can:

• invent cases;
• misquote statutes;
• blend jurisdictions;
• misunderstand procedural posture;
• overstate conclusions;
• ignore exceptions;
• omit implementation steps;
• or generate legal-sounding nonsense.

This is not hypothetical.

Courts have already sanctioned attorneys for submitting AI-generated legal authorities that did not exist.

The American Bar Association addressed this directly in Formal Opinion 512, issued in 2024. The ABA stated that lawyers using generative AI must still consider duties of competence, confidentiality, client communication, supervision, candor to tribunals, and reasonable fees.

The California State Bar has also issued practical guidance on generative AI, emphasizing that lawyers must understand both the benefits and the limitations of AI tools used in legal practice.

The core principle is simple:

Technology may assist the lawyer.

But the lawyer remains responsible.

---

Recent AI Hallucination Problems Are Not Academic

This is no longer a theoretical technology discussion.

Legal AI mistakes are appearing in real cases.

Recent reporting has described lawyers being sanctioned for AI-generated hallucinated citations, including fabricated cases and misrepresented legal authorities. A recent San Francisco Chronicle report noted a growing number of matters affected by AI-generated legal errors and emphasized that courts continue placing responsibility on lawyers and judges—not on the software.

Reuters has also reported on legislative and regulatory responses, including California's SB 574, which was passed by the California Senate and would require attorneys to verify AI-generated materials used in court filings and restrict the input of confidential or personally identifying information into public AI tools.

The lesson is not that attorneys should never use AI.

The lesson is that AI output must be verified.

Every time.

---

The Blind Spot Nobody Talks About: Published Cases Are Not the Entire Story

Another major limitation is more subtle.

Many AI systems depend heavily on public information, published cases, public databases, internet content, and accessible commentary.

That sounds useful.

But it is incomplete.

Some of the most valuable legal lessons never appear in published appellate opinions.

Many real-world outcomes occur through:

• confidential settlements;
• private negotiations;
• unpublished rulings;
• administrative resolution;
• tax structuring decisions;
• implementation experience;
• trustee behavior;
• family conflict management;
• and decades of professional pattern recognition.

AI generally cannot access sealed matters, privileged strategy, confidential negotiations, unpublished tactical outcomes, or nuanced implementation history.

Experienced attorneys can draw from practical judgment that never appears in a public dataset.

That matters because law is not just what appears in a database.

Law is also how rules operate under pressure.

---

Why Estate Planning Is Especially Vulnerable to AI Oversimplification

Estate Planning is not simply:

“Create a trust.”

It involves:

• tax coordination;
• family dynamics;
• beneficiary psychology;
• asset titling;
• business succession;
• fiduciary governance;
• creditor exposure;
• retirement assets;
• community property;
• insurance;
• real estate;
• incapacity planning;
• trust administration;
• and long-term implementation.

A chatbot may generate a trust.

But that does not mean the plan works.

A document can look complete and still fail.

---

Example: “Should I Put My House Into an LLC?”

A client may ask AI:

“Should I transfer my home into an LLC for asset protection?”

The AI may answer:

“Yes.”

But in California, that question can implicate:

• Proposition 19 reassessment issues;
• lender due-on-sale provisions;
• title insurance;
• homeowner's insurance;
• homestead protection;
• transfer tax;
• community property;
• estate tax inclusion;
• capital gains planning;
• and practical financing concerns.

The better answer is usually:

“It depends—and the structure must be evaluated in context.”

That is not a weakness.

That is legal judgment.

---

Example: AI-Generated Trusts and the Funding Problem

A client may arrive with an AI-generated revocable trust and believe planning is finished.

But upon review, the real problems may be:

• the house was never deeded into the trust;
• bank accounts were never retitled;
• brokerage accounts still name individuals;
• retirement beneficiaries conflict with the plan;
• the successor trustee provision is unclear;
• incapacity provisions are weak;
• or distribution terms do not match the family's real objectives.

The client sees a document.

The lawyer sees an unfinished system.

That is the difference.

---

Example: Dynasty Trust Language Without Strategy

AI may be able to generate language that sounds like a dynasty trust.

But dynasty trust planning requires far more than a long document.

It may involve:

• generation-skipping transfer tax planning;
• state situs selection;
• trustee selection;
• distribution standards;
• asset protection objectives;
• beneficiary control concerns;
• income tax treatment;
• decanting flexibility;
• trust protector powers;
• investment governance;
• and family culture.

A dynasty trust is not just a paragraph.

It is a multigenerational governance system.

---

The Rise of AI-Induced False Completion

One of the most dangerous problems emerging in legal practice is what might be called:

AI-Induced False Completion

Clients increasingly believe:

“I already handled my estate plan.”

Why?

Because documents exist.

But documents are not the same thing as planning.

When an attorney reviews the situation, the real picture may be very different:

• trusts were never funded;
• deeds were never recorded;
• LLCs were never maintained;
• operating agreements are inconsistent;
• beneficiary designations contradict the trust;
• powers of attorney are outdated;
• trust provisions do not reflect current tax law;
• successor trustees are inappropriate;
• family conflict was never addressed;
• and assets are still exposed to probate.

AI creates the appearance of progress.

But appearance is not implementation.

Implementation is where estate plans succeed or fail.

---

AI and Copyright: The Regurgitation Problem

Another issue is copyright and proprietary drafting.

Many AI systems have been trained on massive amounts of data.

This raises important questions:

• Was copyrighted material included in training?
• Were proprietary legal forms included?
• Could AI reproduce protected drafting language?
• Could a lawyer unknowingly use language derived from someone else's proprietary work?
• Could client-facing content unintentionally resemble another firm's work?

These questions are still being litigated and debated.

California's Generative Artificial Intelligence: Training Data Transparency Act, effective January 1, 2026, requires developers of certain public-facing generative AI systems to disclose high-level summaries of datasets used in training, including sources, types of data, intellectual property status, and related information. Reuters has reported on litigation involving the Act and trade secret objections to compelled training-data disclosures.

For lawyers, the practical rule is simple:

Do not assume AI-generated content is original merely because it appeared instantly.

Verify it.

Customize it.

Use judgment.

---

Terms of Service May Be the Most Ignored Legal Document in the AI Era

This issue receives far too little attention.

Many professionals upload sensitive information into AI platforms without reviewing:

• terms of service;
• privacy policies;
• data retention rules;
• training permissions;
• confidentiality provisions;
• vendor disclosures;
• system limitations;
• and data cards.

That is risky.

Some tools may store prompts.

Some may use user inputs for training.

Some may retain uploaded documents.

Some may permit broader internal access than users expect.

Attorneys have duties involving competence and confidentiality. ABA Formal Opinion 512 specifically discusses confidentiality, communication, supervision, candor, and fees in connection with generative AI use.

California's practical guidance also emphasizes that AI-enabled tools are increasingly embedded in legal research platforms, document review systems, practice management tools, and other technologies already used by lawyers.

The practical takeaway:

Read the terms.

Review the data policy.

Check whether client information is retained.

Check whether prompts are used for training.

And keep checking.

What was true six months ago may not be true today.

---

Legal AI Hallucinations Are More Common Than Many People Realize

Research confirms that hallucination risk is not limited to casual consumer chatbots.

Stanford researchers evaluated leading legal AI research tools and found measurable hallucination rates even in specialized systems. Stanford's Human-Centered Artificial Intelligence group reported that earlier general-purpose chatbot studies found hallucination rates of 58% to 82% on legal queries, while specialized legal research tools still hallucinated in a meaningful percentage of tested queries.

A Stanford-associated study titled “Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools” examined AI-driven legal research tools offered by major legal research providers and compared them with GPT-4.

The point is not that AI is useless.

The point is that AI must be supervised.

Even when the tool is marketed for legal work.

---

AI Tools Attorneys Are Actually Using

Attorneys are already using AI across many parts of practice.

Common categories include:

Legal Research Tools

These may include AI-assisted versions of major research platforms such as Westlaw, Lexis, Practical Law, and other research systems.

They can help summarize cases, identify authorities, compare standards, and produce first-pass research answers.

But the output must still be verified.

Drafting Tools

AI can help with first drafts of:

• letters;
• memoranda;
• pleadings;
• contracts;
• trust summaries;
• client education pieces;
• intake questionnaires;
• and marketing content.

But drafting is not legal judgment.

Document Review Tools

AI can help review large volumes of documents, flag clauses, identify inconsistencies, and summarize patterns.

This is useful.

But it does not replace attorney review.

Intake and Transcript Tools

AI can summarize client calls, extract action items, identify factual issues, and help organize intake notes.

But attorneys must consider consent, confidentiality, storage, and whether the tool is appropriate for privileged communications.

The New York City Bar has specifically addressed ethical issues involving AI tools used to record, transcribe, and summarize conversations with clients.

Practice Management and Marketing Tools

AI can help create content calendars, blog drafts, social posts, workflows, checklists, and client communication templates.

That can be valuable.

But public-facing legal content must still be accurate, jurisdictionally appropriate, and not misleading.

---

The Client-Side Problem: AI Makes Everyone Feel Like an Expert

Clients now arrive more informed than ever.

That can be good.

But they may also arrive with false confidence.

They may say:

“AI told me I need a Nevada trust.”

“AI said I can avoid taxes with an offshore structure.”

“AI said my LLC protects everything.”

“AI said I can use a dynasty trust and never pay estate tax.”

“AI said I can just amend my trust myself.”

The problem is not that clients are asking questions.

The problem is that AI may have framed the issue incorrectly from the start.

A lawyer then has to unwind not only the legal issue, but also the client's misplaced certainty.

That is why the intake process matters.

The right question is not:

“What document do you want?”

The right question is:

“What exposure are we trying to map, control, reduce, transfer, or eliminate?”

---

Why This Matters for California Estate Planning

California planning is especially difficult to automate because so many issues interact.

A single planning decision may touch:

• California Probate Code;
• California community property law;
• Proposition 19;
• property tax reassessment;
• trust funding;
• homestead exemption concerns;
• creditor protection rules;
• California Rules of Professional Conduct;
• federal estate tax;
• income tax basis planning;
• retirement account rules;
• and family governance.

This is why generic AI answers can be dangerous.

They may be generally plausible and locally wrong.

In legal planning, “almost right” can be expensive.

---

The Difference Between Information and Architecture

AI provides information.

Attorneys provide architecture.

Information says:

“You can create a trust.”

Architecture asks:

• Who controls it?
• What assets go inside?
• What happens at incapacity?
• What happens at death?
• What happens if a beneficiary divorces?
• What happens if a child has creditors?
• What happens if real property is reassessed?
• What happens if tax law changes?
• What happens if trustees disagree?
• What happens if the plan is never funded?

That is the difference between documents and planning.

---

Where AI Actually Helps

AI should not be dismissed.

Used properly, it can be powerful.

AI can help attorneys:

• summarize long transcripts;
• organize client facts;
• prepare checklists;
• draft first-pass educational content;
• generate issue lists;
• create workflow templates;
• compare planning alternatives;
• and improve client communication.

But AI works best as an assistant.

Not as the architect.

Not as the lawyer.

Not as the final decision-maker.

---

Where AI Fails

AI fails when users treat it as a substitute for legal judgment.

It fails when it is used to:

• finalize legal conclusions without review;
• draft complex instruments without attorney analysis;
• answer jurisdiction-specific questions generically;
• handle confidential information casually;
• create tax strategies without tax counsel;
• or produce legal advice without accountability.

AI can accelerate work.

It can also accelerate mistakes.

---

The Future of Law Firms: Commodity Versus Strategic Advisory

Artificial intelligence will likely divide the legal profession into two categories.

Commodity Document Firms

These firms compete on price.

They sell forms.

They automate documents.

They treat law as production.

AI threatens them directly.

Strategic Advisory Firms

These firms solve high-consequence problems.

They focus on:

• wealth transfer;
• asset protection;
• tax exposure;
• business succession;
• family governance;
• trust administration;
• implementation;
• and long-term control.

They do not merely draft.

They architect.

Ironically, AI may make strategic firms more valuable.

The more generic documents become commoditized, the more valuable judgment becomes.

---

Why Human Judgment Still Matters

Legal planning is not only technical.

It is human.

Families are complicated.

Assets are complicated.

Tax laws are complicated.

Trustees make mistakes.

Beneficiaries fight.

Documents get ignored.

Property gets mistitled.

Clients misunderstand funding.

Tax consequences appear years later.

And litigation often emerges from the gap between what the document says and what the family thought it meant.

AI can summarize.

AI can draft.

AI can organize.

But AI does not sit across the table from a family and detect the unspoken conflict.

AI does not understand the hesitation in a spouse's answer.

AI does not know that a beneficiary's addiction, divorce, business risk, or lawsuit exposure changes the entire plan.

AI does not carry malpractice responsibility.

The attorney does.

That is why human judgment still matters.

---

Final Thoughts

Artificial intelligence is extraordinarily powerful.

Used responsibly, it can improve legal practice.

Used carelessly, it can create:

• hallucinated law;
• false confidence;
• confidentiality breaches;
• copyright problems;
• client confusion;
• ethical exposure;
• and implementation failure.

The future does not belong to lawyers who ignore AI.

It also does not belong to lawyers who surrender judgment to it.

The future belongs to professionals who understand:

where AI helps,

where AI fails,

and where human judgment remains irreplaceable.

Estate Planning, Asset Protection, Tax Planning, Business Succession, and family wealth planning are not merely drafting exercises.

They are exercises in judgment under complexity.

And that still requires a human lawyer.

---

Frequently Asked Questions

Can AI create a legally valid trust?

Possibly. AI may generate language that looks like a trust. But validity is not the same as effectiveness. A trust must be properly designed, signed, funded, administered, and coordinated with the client's assets, family structure, tax situation, and long-term goals.

Can AI replace an estate planning attorney?

Not for sophisticated planning. AI may help create drafts or summaries, but Estate Planning requires legal judgment, implementation, tax awareness, fiduciary design, family dynamics, and asset coordination.

Why are AI hallucinations dangerous in law?

AI hallucinations are dangerous because the output can sound confident even when it is wrong. In law, a fabricated case, wrong statute, or missing exception can create serious legal consequences.

Can AI access unpublished cases?

Generally, AI tools rely heavily on available public data, published authority, and licensed databases. They usually do not have access to sealed matters, confidential settlements, privileged strategy, or private implementation experience.

Should lawyers upload client information into AI platforms?

Only after carefully reviewing the platform's terms of service, privacy policy, confidentiality protections, data retention rules, and training practices. Lawyers must protect confidential client information.

Can AI-generated legal content contain copyrighted material?

Potentially. There are ongoing disputes and policy debates over training data, copyrighted works, and whether AI can reproduce proprietary language. Lawyers should verify and customize AI-generated material.

What is ABA Formal Opinion 512?

ABA Formal Opinion 512 is the American Bar Association's 2024 ethics opinion addressing lawyers' use of generative AI. It discusses duties including competence, confidentiality, communication, supervision, candor, and reasonable fees.

Is AI useful in legal practice?

Yes. AI can be useful for summaries, checklists, workflow support, drafting assistance, research organization, and client education. But it should assist legal judgment, not replace it.

What is AI-induced false completion?

AI-induced false completion occurs when clients believe their legal planning is finished because AI generated documents, even though the plan has not been reviewed, funded, coordinated, or implemented properly.

What is the safest way to think about AI in law?

Think of AI as an assistant, not the attorney. It may help organize information and create drafts, but legal judgment, verification, strategy, and responsibility remain human obligations.

---

Authorities, References, and Resources

American Bar Association, Formal Opinion 512, Generative Artificial Intelligence Tools, issued July 29, 2024. The opinion addresses competence, confidentiality, communication, supervision, candor toward tribunals, and reasonable fees for lawyers using generative AI.

State Bar of California, Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law. The guidance explains that AI tools are increasingly embedded in legal research platforms, document review systems, and practice management tools, and clarifies how professional responsibility obligations apply.

State Bar of California, Ethics & Technology Resources. The toolkit addresses generative AI opportunities, risks, confidentiality, bias, accountability, and best practices for lawyers.

New York City Bar Association, Formal Opinion 2025-6, Ethical Issues Affecting Use of AI to Record, Transcribe, and Summarize Conversations With Clients. The opinion discusses ethical obligations arising when attorneys use AI tools in client communications.

Stanford Human-Centered Artificial Intelligence, “AI on Trial: Legal Models Hallucinate in 1 out of 6 or More Benchmarking Queries,” May 23, 2024. Stanford reported substantial hallucination rates in legal AI tools and even higher rates in general-purpose chatbot legal queries.

Magesh et al., “Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools,” evaluating AI-driven legal research tools and comparing them with GPT-4.

San Francisco Chronicle, report on California lawyers, AI hallucinated citations, and judicial sanctions, published May 2026.

Reuters, report on California SB 574 regulating lawyers' use of AI, including verification of AI-generated materials and restrictions on confidential or personally identifying information in public AI tools.

Reuters, report on California's Generative Artificial Intelligence: Training Data Transparency Act and litigation involving AI training data disclosures and trade secret objections.

ABA Model Rules of Professional Conduct, including Rule 1.1 Competence, Rule 1.6 Confidentiality, Rule 3.3 Candor Toward the Tribunal, Rule 5.1 Responsibilities of Supervisory Lawyers, and Rule 5.3 Responsibilities Regarding Nonlawyer Assistance.

California Rules of Professional Conduct, including competence, confidentiality, supervision, communication, candor, and duties related to client information and professional responsibility.

California Business and Professions Code § 6068, including attorney duties of confidentiality, honesty, and faithful discharge of professional obligations.

Authorities, References, and Resources

The following authorities and resources were reviewed or are relevant to the discussion above:

• American Bar Association — Formal Opinion 512, Generative Artificial Intelligence Tools (2024). This opinion addresses attorney duties involving competence, confidentiality, communication, supervision, candor, and fees when using generative AI.
• New York City Bar Association — Formal Opinion 2024-5, Generative AI in the Practice of Law. This opinion provides guidance on ethical obligations when lawyers and law firms use generative AI in legal practice.
• Reuters — Lawyers using AI must heed ethics rules, ABA says in first formal guidance.
• Stanford Human-Centered AI — AI on Trial: Legal Models Hallucinate in 1 out of 6 or More Benchmarking Queries.
• Stanford/Legal AI Research Study — Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools.
• Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models.
• ABA Model Rules of Professional Conduct:
  • Rule 1.1 — Competence
  • Rule 1.6 — Confidentiality
  • Rule 3.3 — Candor Toward the Tribunal
  • Rule 5.1 — Responsibilities of Supervisory Lawyers
  • Rule 5.3 — Responsibilities Regarding Nonlawyer Assistance
• California Rules of Professional Conduct and California Business and Professions Code § 6068, including duties involving competence, confidentiality, honesty, supervision, and faithful discharge of professional obligations.

---

Legal Disclaimer

This article is provided for general educational and informational purposes only. It is not legal advice and should not be relied upon as legal advice for any particular person, family, business, trust, estate, transaction, or legal matter.

Reading this article does not create an attorney-client relationship with the Law Office of James Burns or James G. Burns, Esq. Legal planning depends upon the specific facts, documents, assets, family circumstances, tax issues, jurisdictional rules, and implementation requirements involved.

Anyone considering the use of artificial intelligence in legal planning, estate planning, asset protection, business succession, tax planning, or trust administration should consult qualified legal counsel before acting.

---

Intellectual Property Disclosure

This article, including its structure, phrasing, strategic framework, examples, and explanatory approach, is the intellectual property of the Law Office of James Burns unless otherwise noted.

No portion of this article may be copied, reproduced, republished, modified, distributed, or used for commercial purposes without prior written permission, except for brief quotations with proper attribution.

References to third-party publications, ethics opinions, studies, articles, or authorities are included for educational purposes and remain the property of their respective authors, publishers, and organizations.