A confidently stated AI output that is false or erroneous. It may include invented or misapplied legal information.

Identifying questions, missing facts, deadlines, or topics that may need further checking. It is not a final legal conclusion.

Check generated facts against original documents and check legal authorities, dates, and citations against current authoritative sources.

A federal procedural rule that imposes responsibilities concerning representations made to a court.

What is confabulation?

Confidently presented AI content that is erroneous or false.

What is issue spotting?

Identifying legal questions, missing facts, or possible areas for further research without deciding the outcome.

What is professional judgment?

The accountable application of legal knowledge, facts, ethics, and a person's objectives to a specific matter.

What AI Can—and Cannot—Do with a Legal Question — AI as a Legal Assistant—not an Attorney

1. Start with the right division of labor

Two layers of a legal question

Legal problems involve information work and professional judgment. AI can help organize and explain information; legal judgment decides what to do with it.

Why the boundary matters

AI has no license, client relationship, duty of loyalty, or authority to choose a legal strategy for you. It can assist, but it cannot represent you.

2. Understand why fluent answers can fail

Fluency is not evidence

AI predicts plausible language. It may therefore give a detailed answer that sounds authoritative without being legally accurate, current, or applicable.

Confabulation

NIST uses confabulation for confidently stated false or erroneous output. Treat every AI-provided authority, date, and quotation as a claim to verify.

3. Use AI for plain-language orientation

Turn a document into a map

For a lease notice, AI can restate the text, identify terms, extract dates, and generate questions. That can make a stressful document easier to inspect.

Do not outsource the decision

Whether to respond, pay, negotiate, or challenge a notice depends on the actual document, governing law, evidence, and consequences—not a generic chatbot answer.

4. Use AI to organize, not to decide what is true

Organization is high-value support

AI can group records and create a timeline of events, sources, and missing information. This helps people see a dispute as a structured set of facts.

A summary is not the record

AI cannot decide credibility, authenticate evidence, or resolve conflicts between documents. Check the original material before relying on any generated summary.

5. Drafting is a starting point, not a finished legal instrument

Draft narrowly

Ask AI for a timeline, question list, or neutral first draft. Narrow prompts reduce the chance that it fills gaps with assumptions or unsupported conclusions.

Verify before use

Check facts against originals and legal authorities against official or reliable primary sources. A court submission carries special responsibilities for accuracy.

6. Protect confidential and sensitive information

Prompts can expose information

A legal prompt may contain sensitive personal, financial, medical, business, or privileged material. Sharing it is a data-handling decision, not merely a writing task.

Minimize what you share

Remove identifiers and unnecessary documents. Learn how a tool handles inputs, storage, access, and retention before using it with sensitive legal material.

7. Try the task-boundary test

Sort the request

Read each request and decide whether it is primarily assist, verify, or professional judgment.

"Put these dated emails into a timeline." -> Assist.
"Is this court deadline still valid in my jurisdiction?" -> Verify against the current official rule or docket.
"Should I accept this settlement offer?" -> Professional judgment.

Now test your own legal question. Circle the words that signal a decision: `should`, `must`, `best`, `risk`, `likely outcome`, or `file`. Those are clues that AI output should not be the final authority.

8. Review the key terms

Flip the cards to review the limits and useful roles of AI in legal questions.

Confabulation: A confidently stated AI output that is false or erroneous. It may include invented or misapplied legal information.
Issue spotting: Identifying questions, missing facts, deadlines, or topics that may need further checking. It is not a final legal conclusion.
Professional judgment: A lawyer's responsibility to interpret law and facts, advise a client, make strategy choices, and remain accountable for the work.
Verify-before-use: Check generated facts against original documents and check legal authorities, dates, and citations against current authoritative sources.

Check understanding

Choose the safest use of AI when preparing to address a legal problem.

Which request is the best fit for AI assistance before human review?

Create a dated timeline from verified emails and identify missing documents
Choose whether to accept a settlement based on the likely court outcome
Submit a court filing with AI-generated citations without checking them

Show Answer

Answer: A) Create a dated timeline from verified emails and identify missing documents

Organizing verified records and identifying gaps is a support task. Settlement decisions require contextual professional judgment, and all court-facing claims and citations must be independently checked.

Key Terms

Rule 11: A federal procedural rule that imposes responsibilities concerning representations made to a court.
confabulation: Confidently presented AI content that is erroneous or false.
issue spotting: Identifying legal questions, missing facts, or possible areas for further research without deciding the outcome.
professional judgment: The accountable application of legal knowledge, facts, ethics, and a person's objectives to a specific matter.