ABA Formal Opinion 512: What Law Students Should Know About Lawyers Using AI

ABA Formal Opinion 512 is not an anti-AI document.

It is a lawyer-responsibility document.

The ABA's July 29, 2024 Formal Opinion 512 applies existing duties around competence, confidentiality, communication, supervision, candor, and reasonable fees to generative AI.

For law students, the takeaway is simple: AI use is becoming normal, but lawyer accountability does not move to the software.

Competence still means understanding the tool

A lawyer does not need to become a machine-learning engineer to use a legal AI product. But the lawyer does need a reasonable understanding of what the tool can and cannot do.

That means knowing whether the tool:

• Uses public model outputs or legal-source-grounded results
• Can hallucinate citations
• Stores prompts or documents
• Trains on user input
• Can access client documents securely
• Produces answers that need human legal review

For a junior associate, this changes the job. You may be asked to use AI to accelerate research, summarize documents, or draft first passes. You still need to verify the work. The person who signs the filing, sends the memo, or advises the client cannot blame the model for a bad answer.

Confidentiality is the first practical issue

The fastest way to misuse AI in legal work is to paste confidential client material into a tool that is not approved for that use.

Law firms will increasingly have internal AI policies, approved tools, and matter-specific rules. Follow them. If you are a summer associate, do not improvise. If you are unsure whether a tool is approved for client documents, assume it is not until someone tells you otherwise.

This is where legal AI differs from using a chatbot to plan a vacation. Client documents, deal terms, witness names, deposition transcripts, and litigation strategy are sensitive. The convenience of a tool does not override confidentiality.

Client communication can matter

Formal Opinion 512 points to the duty to communicate with clients. That does not mean every use of AI requires a dramatic disclosure. It does mean some uses may require client communication or informed consent, depending on the work, the tool, the client's instructions, and the jurisdiction.

Students should understand the distinction.

Using an approved internal tool to organize a document set may be different from using an outside tool to generate legal analysis from confidential materials. Using AI for administrative help may be different from using it in a way that affects strategy, fees, or client expectations.