Field Note: 21 Ways AI Can Hallucinate in Your Legal Brief

In this Field Note, Ron Drescher breaks down one of the most important—and misunderstood—risks in legal AI: hallucinations.

The episode begins with the recent Sullivan & Cromwell filing admitting AI-generated errors, with a close look at the now-famous Schedule A. While most commentary has focused on fake citations and misquotes, Ron highlights the more subtle—and more dangerous—types of hallucinations that appeared in that filing.

From structurally corrupted citations to mutated judicial language, this episode explores how AI doesn’t just make obvious mistakes—it makes mistakes that look like law.

Ron then expands the discussion to a broader framework, identifying both the most well-known hallucination risks and the lesser-known categories that are more likely to survive verification and make their way into filed briefs.

⚖️ What You’ll Learn

• Why the Sullivan & Cromwell Schedule A is more important than the confession letter

• Two underappreciated hallucinations:

• Citation drift (hybrid citation corruption)

• Mutated quotations

• The 3 most common AI hallucinations:

• Fabricated cases

• Real cases with incorrect holdings

• Invented quotations

• Three lesser-known (and more dangerous) hallucinations:

• Subtle semantic drift

• Fake multi-case consensus

• Logical hallucination (broken arguments that look complete)

• Why “just verify the citation” is no longer enough

• A practical verification framework for AI-assisted legal writing

🧠 Key Takeaway

AI hallucinations are no longer edge cases—they are part of the operating environment of modern legal writing.

The real risk isn’t obvious errors.

It’s the errors that:

look correct

pass a quick check

and still make it into your brief

📥 Downloadable Resource

This episode includes a companion Field Note:

👉 “21 Ways AI Can Hallucinate in Your Legal Brief”

Use it as a working reference during your hallucination verification process—not as a one-time read.

🔧 The New Verification Standard

Before including any authority in a brief, confirm:

Does the case support the proposition?

Is the quote accurate and in context?

Does the procedural posture match your argument?

Has the legal standard shifted subtly?

🔜 Coming Next

Field Note:

12 Ways BigLaw Associates Are Quietly Optimizing AI in Legal Drafting

A practical look at how lawyers in high-stakes environments are adapting their workflows to use AI effectively—without getting buried in verification.

About the Show

AI Tools for Practicing Lawyers delivers practical, no-nonsense guidance to help attorneys use AI safely, effectively, and profitably in the modern practice of law.