The Myth of AI Discoverability

First a disclaimer about using AI for legal work

AI is no substitute for actual legal research, and it cannot replace the judgment of an attorney. Alabama law is too complex for that. Courts frequently interpret statutes in ways that diverge from their (sometimes conflicting) plain text, and outcomes often hinge on context, precedent, and procedure rather than the surface meaning of a statute. Because of this, AI use should be treated as a discussion rather than authorityit is a starting tool. Even when I use AI to proof a final draft of an email I make sure it doesn't change anything.

AI Anxiety

People worry that AI research might be discoverable or used against them. This concern is understandable in a digital era where information feels permanent and where missteps with technology occasionally make headlines. However, it seems the fear exceeds the reality. AI research sessions resemble private brainstorming or handwritten notes. They are not automatically preserved, not automatically shared, and not automatically accessible to opposing parties. Discovery depends on relevance, possession, and intentional preservation. If an attorney does not save or transmit AI‑generated content, there is nothing to produce. Don't believe the conspiracy theories.

Your AI Research Does Not Create Automatic Exposure

Courts distinguish between evidence and the internal thought processes of counsel. Mental impressions, preliminary analyses, and exploratory research fall into a protected zone that is not subject to routine discovery. AI‑assisted research fits comfortably within that category. It is simply another medium through which an attorney thinks through a problem. Privilege is not waived by private research, and confidentiality is not compromised unless a user voluntarily inputs sensitive client information into a public system. The risk lies not in the act of researching but in the careless handling of privileged facts.

The Risk: Misunderstanding What AI Can and Cannot Do

The deeper concern is not discoverability but misunderstanding of teh value of AI products. AI can give a surface-level summarization and synthesize given pre-written information, but it cannot interpret precedent, weigh competing authorities, or anticipate judicial reasoning or a judge's desire to be fair to an unrepresented person or a person who is on the losing end becasue they were ignorant.

AI cannot replicate the human judgment required to understand why a court reached a particular conclusion in a specific case or how a judge might apply a doctrine in a given fact pattern - but it will try. When attorneys treat AI as a unexperienced assistant rather than an oracle, the tool becomes safe, efficient, and reliable.

Regarding AI Discoverability

In federal court, the Western District of Arkansas ruled in Whisman v. Rajabi that a pro se plaintiff’s chats with Microsoft Copilot were not discoverable emphasized that using Copilot did not waive work‑product protection because disclosure must be to an adversary or someone likely to pass information to an adversary — and “AI programs are tools, not people.”

Conclusion

AI research is not a trapdoor into discovery, nor a mechanism through which opposing counsel can peer into your strategy. It is a modern extension of traditional research methods, valuable when used wisely, and harmless when used with discretion. Attorneys who verify AI‑generated results through

primary sources and avoid inputting privileged information can rely on these tools without fear. The law remains a human endeavor, and AI remains exactly what it should be: a helpful, fast, but ultimately subordinate assistant in the hands of a trained professional.*

• This article was generated with AI help, so any mistakes or tyopis are the AI's fault, not mine. ;-)

Gregory Stanley is an Alabama Land Lawyer who has handled hundreds of real property cases for investors in Alabama. Do not take legal advice from the internet.