Lana Akopyan

Judgment

Aug 27, 2026

•

3 min read

The Expiring Audit

An assessment can become stale without ever having been wrong.

Lana Akopyan

Judgment

Aug 21, 2026

•

4 min read

The Instruction You Cannot See

A court just identified the half of AI review that human verification does not reach.

Lana Akopyan

Judgment

I Corrected the Paragraph Fifteen Times. It Was Still Wrong.

Aug 10, 2026

•

3 min read

I Corrected the Paragraph Fifteen Times. It Was Still Wrong.

Every edit was justified. The paragraph was built around the wrong point.

Lana Akopyan

Benchmarks

The Number Was Right. The Conclusion Wasn't.

Aug 3, 2026

•

4 min read

The Number Was Right. The Conclusion Wasn't.

How a benchmark can stay accurate while becoming misleading.

Lana Akopyan

Benchmarks

“A Human Reviewed It.” What Does That Actually Prove?

Jul 30, 2026

•

4 min read

“A Human Reviewed It.” What Does That Actually Prove?

Human review matters. It is also being asked to prove things it cannot establish.

Lana Akopyan

Confidentiality

Your AI Stopped Answering and Started Acting

Jul 19, 2026

•

3 min read

Your AI Stopped Answering and Started Acting

Once the tool can read files, send messages, and call outside services, the unit of confidentiality risk is no longer the prompt. It is the access.

Lana Akopyan

Confidentiality

Jul 14, 2026

•

4 min read

A Thumbs-Up Can Last Five Years

Rating a Claude or ChatGPT answer feels like grading the tool. On Anthropic's commercial products, that click is permission to keep the whole conversation and train on it. On OpenAI's, it depends on which product you are in and a setting most lawyers have never seen. The same click can be a disclosure decision - and the client was never asked.

Lana Akopyan

Confidentiality

Your AI Vendor Keeps a Copy. That Is Not the Problem.

Jul 9, 2026

•

4 min read

Your AI Vendor Keeps a Copy. That Is Not the Problem.

An AI vendor's copy of a client trade secret, held under confidentiality terms, does not lose the secret. What loses it is an obligation that expires while the copy lives on, or a copy you cannot require the vendor to delete.

Lana Akopyan

Confidentiality

Jun 30, 2026

•

4 min read

No Training Does Not Mean No Retention

Every enterprise AI tier promises not to train on your data. None of them promises not to keep it. Lawyers are treating the first as if it were the second.

Lana Akopyan

Confidentiality

Jun 23, 2026

•

4 min read

Deletion Is Not Deletion

You can delete your chat history. You cannot delete the copy the provider kept - and that copy, while it lasts, is the one legal process can reach.

Lana Akopyan

Confidentiality

AI Did Not Break Privilege. There Was Never Privilege to Break.

Jun 18, 2026

•

5 min read

AI Did Not Break Privilege. There Was Never Privilege to Break.

Everyone is reading United States v. Heppner as the case where a chatbot destroyed attorney-client protection. Read it beside a quieter ruling from the same weeks that came out the other way, and the real lesson is one lawyers already knew.

Lana Akopyan

Confidentiality

Jun 16, 2026

•

3 min read

Which Claude Are You Actually Using?

There is no single answer to what the tool does with your inputs. The terms depend on how you reach the model and which one you pick - and they changed three times while you were not looking.

Lana Akopyan

Confidentiality

An AI Model Disappeared Overnight. The Lesson Was Already True.

Jun 13, 2026

•

3 min read

An AI Model Disappeared Overnight. The Lesson Was Already True.

What the Fable shutdown actually teaches lawyers - and it is not the thing everyone is arguing about.

Lana Akopyan

Judgment

Jun 10, 2026

•

2 min read

If You Used AI, Can You Prove Judgment Survived?

Competence is shifting from "is the output correct" to "can you show where your judgment took control." A short record closes most of the gap.

Lana Akopyan

Confidentiality

A Prompt Can Reveal a Client Without Naming Them.

Jun 9, 2026

•

3 min read

A Prompt Can Reveal a Client Without Naming Them.

For IP and patent lawyers especially, "redact the client's name" is not a confidentiality strategy. It barely touches the risk.

Lana Akopyan

Judgment

Jun 8, 2026

•

3 min read

Same Failure, Better Suit.

The AI mistakes that get pro se litigants mocked are happening inside expert legal work too. They are just better dressed, which is what makes them dangerous.

Lana Akopyan

Judgment

AI Is Built to Agree With You. That Is the Problem.

Jun 7, 2026

•

3 min read

AI Is Built to Agree With You. That Is the Problem.

Why legal competence now requires lawyers to preserve friction, not just verify outputs.

Lana Akopyan