By Chuck Gallagher — Business Ethics Keynote Speaker and Trainer
TL;DR: Chuck Gallagher, an AI ethics speaker and author, breaks down three ways artificial intelligence is upending white-collar investigations: evidence you cannot trust, chatbot prompts that may not be private, and government searches that reach further than ever.
How AI Is Changing White-Collar Investigations
Picture a conference room. A lawyer plays a thirty-second clip. In it, an executive tells a subordinate to move a number before the quarter closes. The voice is right. The face is right. The timestamp checks out.
It never happened.
That scene isn’t fiction. It’s what investigators, prosecutors, and defense counsel are staring at right now, all across the country. Writing in the New York Law Journal this month, white-collar attorney Brendan Quigley laid out three ways artificial intelligence is forcing everyone in a criminal investigation to rethink assumptions the law has leaned on for decades. Whether the evidence is real. Whether a conversation with a machine is protected. And how far a government search can now reach. None of these is a someday problem. Courts are wrestling with all three today.
I want to walk through each one, because underneath all that legal machinery sits a very old human story. Every tool that makes deception easier also makes trust more expensive. As an AI ethics speaker and author, I’ve watched that pattern repeat across every technology I’ve studied. AI just raises the stakes.
Can You Still Trust What You See?
Start with authentication. For most of legal history, a photograph or a recording carried a kind of built-in credibility. Seeing was close to believing. That’s over. Now a convincing fake video or a cloned voice can be produced by anybody with a laptop and a little patience. Courts are beginning to confront this head-on, asking how AI-generated or AI-altered material gets tested before a jury ever lays eyes on it. A judge can’t just take a clip at face value anymore. Somebody has to prove where it came from.
Here’s the twist most people miss. Everyone worries about the fake that gets admitted. The quieter danger is the real recording that gets doubted. When everyone knows fakes exist, a guilty party can wave a hand at genuine evidence and call it manufactured. Lawyers already have a name for this. The liar’s dividend. The mere possibility of forgery pays out to the dishonest, and it charges the honest their credibility.
Is a Chatbot Prompt Privileged?
Second issue: privilege. Think about how many people now type their real questions into an AI tool. An employee uneasy about a deal asks a chatbot to spell out the risk. A manager drafts a sensitive memo with an assistant’s help. Every one of those prompts, and every one of those answers, is a record. It persists. And it can be subpoenaed.
Attorney-client privilege was built for a world of lawyers and letters. Nobody wrote it for a world where the first place you take a legal-flavored worry is a machine owned by some company you’ve never met. Was that prompt a private thought? A business record? A confession typed into a search box? Courts have barely begun to answer. And companies treating their AI logs as harmless housekeeping are in for a hard surprise.
How Far Can a Search Reach Now?
Third: search. Investigators used to be limited by sheer volume. There was only so much a team of humans could read. AI removes that ceiling. Government analysts can now sift oceans of chat messages, prompt logs, and model outputs, then line them up against market events with a precision that didn’t exist a decade ago. Under the Fourth Amendment, that power cuts both ways. How much can a search touch before it turns into a fishing expedition? The doctrine is still catching up to the tools.
Same Sin, New Speed
Now, the way I see it, none of this changes the underlying moral question. What it changes is the speed and the scale. Deputy Attorney General Lisa Monaco said it plainly a couple of years back: fraud using AI is still fraud, price fixing using AI is still price fixing, and market manipulation using AI is still market manipulation. The statute doesn’t care which tool you reached for. Prosecutors have already shown they’ll pick up old laws to answer new tricks. In one Southern District of New York case, they charged a defendant in a digital-asset insider trading scheme using a wire fraud statute written back in the 1950s.
Let me be clear about what all this means for you if you run a company or advise one. Your AI tools aren’t neutral bystanders sitting off to the side. They’re witnesses. The prompt someone typed at eleven at night, the output the model handed back, the log nobody thought to delete. Any of it can end up in a prosecutor’s hands, or on a courtroom’s record. I tell leaders the same thing every time. The technology didn’t create the temptation — it just made the paper trail longer and the lie easier to build.
Every choice has a consequence. That was true when the evidence lived in a filing cabinet. It’s truer now that the evidence is everything you ever typed.
Frequently Asked Questions
What are the three AI issues reshaping white-collar investigations?
Three pressure points keep coming up. Whether digital evidence can still be trusted. Whether your conversations with AI tools are protected by privilege. And how far government searches can now reach through AI-sorted data. Each one pokes at an assumption criminal law has leaned on for decades, and the courts are only starting to work out the answers.
What is the “liar’s dividend”?
It’s the advantage a dishonest person gets for free, just because convincing fakes now exist. Once a jury knows video and audio can be forged, a guilty party can shrug off real evidence as fabricated — and the truth gets harder to prove.
Are my AI chatbot prompts protected by attorney-client privilege?
Usually not. The privilege covers confidential communications with your lawyer. A question typed into a third-party AI tool is a different animal — a record that can be stored, subpoenaed, and read back to you in an investigation. Treat anything you type into a chatbot as potentially discoverable.
Can the government use AI to search my company’s records?
Yes. The scale is what’s new. AI lets investigators sort enormous volumes of messages, logs, and files, then line them up against outside events at a speed no human team could ever match. How the Fourth Amendment limits a search that powerful is something courts are still working out.
Is committing fraud with AI treated differently under the law?
No. The tool doesn’t change the crime — I say this constantly. A former Deputy Attorney General put it simply: fraud committed with AI is still fraud. And prosecutors have shown they’ll reach for existing statutes, some of them decades old, to charge AI-enabled schemes.
A Word Before You Go
The tools are new. But the choice underneath them is the same one people have always faced. That’s the conversation I bring to boardrooms, conferences, and leadership teams wrestling with how AI reshapes ethics, evidence, and accountability. If you want that message at your next event — or you want to build a culture where the paper trail never has to embarrass you — you’ll find me at ChuckGallagher.com.
Five Questions for Reflection
1. If a convincing fake could be made of you saying something you never said, how would you prove it wasn’t real?
2. What have you or your employees typed into an AI tool that you would not want read aloud in a courtroom?
3. Does your organization treat AI prompts and outputs as business records, or as private scratch paper?
4. Where is the line between a lawful search and a fishing expedition when a machine can read everything at once?
5. If the technology only makes dishonesty easier, what does that say about the honesty you already have?
