Bench & Bar May/June 2026

FEATURE: ARTIFICIAL INTELLIGENCE TECH & TRUST:

APPLYING AN IT MINDSET TO AI IN ESTATE PLANNING AND ELDER LAW BY SHELLY ANN KAMEI

L ong before I navigated the labyrinth of the law, I worked in the tech boom of the late 1990s and early 2000s, translating complex systems into practical workflows as a technical writer, trainer, and project manager. That background informs my view of today’s AI gold rush: it isn’t magic or madness. AI is a powerful tool that requires guardrails and human oversight to ensure it serves, rather than undermines, our ethical obligations—particularly in estate planning and elder law, where risks may take years or decades to surface. The most persistent mistake I see in the legal community is what I call the ostrich problem: AI feels overwhelming, so some attorneys try to ignore it. But opting out isn’t possible. If you use the Microsoft, Google, or Apple ecosystems, you are already relying on AI—often in ways you don’t see. You can restrict features, but you cannot avoid AI. The second major mistake is buying a sub scription to a popular AI tool and then asking, “What can I do with this?” That is the opposite of how seasoned IT profession als approach software and systems. MY IT MANTRA IS SIMPLE: “DON’T PICK UP A HAMMER AND SPEND YOUR DAY SEARCHING FOR NAILS.” THE IT VETERAN’S APPROACH TO AI

We begin with the problem: Where is there a bottleneck that could be relieved through surgical use of AI? Once the need is defined, we evaluate all automation and AI tools through Kentucky Legacy Law’s technical sovereignty lens. This is a three-pillar framework as follows: Does the provider contractually and tech nically segregate our data from their global training set? Is the data siloed within my tenant? 1 If not, is the connection secured consistent with our professional obligations? If needed, can I produce a log of every interaction? Do not compromise your ethics for cost, speed, or even a seemingly superior plan. THE MULTI-HEADED HYDRA It’s a mistake to treat AI as an isolated threat. The convergence of multiple tech nologies supercharges the dangers each would pose on its own. AI magnifies these risks exponentially, but it’s only one part of a much larger, interconnected ecosystem. The related risks include: 2 1. THE PERIMETER Much of your firm’s data no longer lives on machines you control inside your office. It

resides in thirdparty storage in the cloud. This means it’s replicated across multiple data centers, subject to vendors’ retention settings, logs, and background services. In practical terms, your confidentiality obliga tions extend beyond your own hardware to every platform that stores, caches, backs up, or indexes your files. If you do not know the terms of service for these custodians of your data, you may be inadvertently breaching your ethical duties. 2. METADATA: SILENT BUT PERSISTENT Modern AI sees what human cannot. It rec ognizes that a PDF is a layered container where residual data survives cosmetic edits. Searchandreplace or black rectangles don’t remove that data; it remains selectable, searchable, and extractable. This happened in the Epstein files release: some PDFs used visual overlays rather than true, destructive redaction, so people could copypaste or otherwise extract the hidden text. In other words, the black bars covered the text, but the text was still inside the file either behind the bars or as metadata. The only safe approach is true redaction ( i.e. , permanently removing content and sanitizing metadata), not visual masking. In my firm, we redact using this method, print the document, and then rescan it. There is no hidden data retained in the document. 3

16 may/june 2026