home  /  security & threats
department of security & threats

A law firm holds other people's money and other people's secrets.

That combination is why firms are targeted disproportionately — and why AI changed the economics of attacking one.

begin here

Where is your firm?

Start a conversation with the AI Adoption Concierge, already scoped to security & threats. Choose the question closest to yours, or describe your situation directly.

AI Adoption Conciergesecurity & threats · orientation, not legal or ethics advice
Tell me the firm's size, whether you hold client funds, and what AI is in use today — sanctioned or not. I'll help you work out where the real exposure is. This isn't a security audit and isn't legal advice.

Most security material written for law firms is generic material with the word "law firm" inserted. The specific exposure is narrower and worse than that. A firm holds client funds in trust and holds pre-announcement commercial secrets for dozens of better-defended organisations, which makes it a concentrated target rather than an incidental one — a point federal advisories aimed at the legal sector have made explicitly. AI has changed three things about that: it made convincing impersonation cheap enough to use against every firm rather than the largest ones; it introduced a genuinely new attack that arrives inside documents lawyers are professionally obliged to read; and it created a category of confidentiality exposure that produces no alert, no ransom note and no moment of discovery. The controls that address the worst of it are unglamorous and mostly free.

specialization areas

Areas in this part of the practice.

What is being used against firms, what defends them, and how to deploy AI without creating the exposure you were trying to avoid.

methodology

How this department investigates.

How the Institute approaches security here — evidence over vendor claims.

Incidents are dated and sourcedThreat claims age faster than anything else in AI. An undated figure is not evidence.
Vendor claims held against measurementWhere independent research disagrees with vendor marketing in this field, it usually disagrees by a wide margin.
Controls scaled to firm sizeThe same control set, differently bought. A five-person firm and a five-hundred-lawyer firm need the same protections and almost none of the same products.
Cost stated plainlyIncluding where the answer is that the highest-value control is free.
Gaps named as gapsWhere the public record is silent, saying so is more useful than inference dressed as fact.
Categories, never rankingsWhat a category does and what to ask of it. The Institute does not rank vendors.
common questions

Security — the questions firms ask.

What is the single highest-value thing a small firm can do?

Write down a wire-verification protocol and enforce it. Callback to a number from the firm's own records — never a number supplied in the request — a mandatory hold on any change to payment instructions, and explicit written authority for any member of staff to delay a transfer to verify without being second-guessed. It costs nothing, and it defeats the attack that has the largest single-event consequence for a firm. Federal reporting on wire fraud in real-estate closings includes matters where the fraudulent instruction impersonated the buyers' own attorneys.

Has a law firm actually lost client data through an AI tool?

No such case could be found in the public record as of September 2026. That is worth saying plainly, and worth not over-reading. Three explanations fit equally well: enterprise-tier deployments have held; it has happened and firms have strong incentives not to disclose a confidentiality failure; or it is happening through unsanctioned consumer-tool use, which produces no alert and no moment of discovery. The absence of a named cautionary case is the main reason firm leadership underweights this, and it is not evidence of safety.

Is AI more useful to attackers or defenders right now?

Attackers, at the margin, because the gains are asymmetric. Controlled research found AI-generated spear-phishing matched a skilled human operator's success rate at a fraction of the cost — which does not make it better than expert phishing, it makes expert-quality phishing affordable against every employee of every firm. On the defensive side, independent measurement of AI security tooling has consistently come in far below vendor claims. The honest position is that AI has raised attack volume and quality more than it has raised defensive capability, and that most of the useful defensive AI is in email security rather than anywhere more exciting.

Does this belong to IT?

Partly, and the parts that do not are the ones that get missed. Deciding which AI tier the firm buys is a confidentiality decision with a procurement mechanism. Deciding whether an agent may act on a matter is a supervision question. Deciding what happens in the first hour after an incident is a professional-responsibility question. Firms that delegate the whole area to IT tend to end up with good endpoint protection and an unanswered question about what their lawyers are pasting into a chatbot on a personal phone.

Not sure where your exposure actually is?

Describe the firm and how AI is being used today. The Institute will help you find the gaps that matter.

AI adoption conciergeorientation · not legal or ethics advice
Tell me the firm's size, whether you hold client funds, and what AI is in use today — sanctioned or not. I'll help you work out where the real exposure is. This isn't a security audit and isn't legal advice.