Deputy Assistant Attorney General Daniel Glad Delivers Remarks at the 2026 Women’s White Collar Defense Association Conference

Remarks as Prepared for Delivery

It is always good for me to be back in Chicago. This is home. I grew up here. I started my career here. I worked in city government and I prosecuted cases here. The Antitrust Division has an office here, just a few blocks away, where I got my start as a federal prosecutor.  Our office on LaSalle Street has done a lot over the years to advance the mission, from bringing a landmark global price-fixing case (that became a Hollywood film) to producing two thirds of the current criminal front office and the chief of staff at the Antitrust Division.

A while ago, I moved from Chicago, to Washington. Since I left, the Bears are winning, the White Sox are pennant contenders, and a Chicagoan became Pope. At some point, you have to consider the possibility that I was the problem.

So, this is a good place to return to a subject I spoke about earlier this year.

Old Crime, New Code: A Reprise & an Update

In May, I gave remarks I called “Old Crime, New Code,” which is what prompted the WWCDA invitation. The basic point I made then was simple: Technology can change how competitors communicate. It can change how prices are generated. It can change where information is stored and who has access to it.

It does not change the basic requirement that competitors make independent competitive decisions. Software cannot launder collusion.

What technology can change is the mechanism through which competitors coordinate. It can change where the evidence resides. It can change who has visibility into that evidence. And the incentives we at DOJ create can change who has a reason to bring that evidence to us.

Let me start briefly with the law. When I spoke in May, I discussed Cornish-Adebiyi v. Caesars Entertainment, a private civil case involving allegations that competing Atlantic City hotels used a common algorithmic pricing platform. The district court had dismissed the complaint because it concluded that the plaintiffs had not adequately alleged an agreement among the competing hotels themselves — the rim of the alleged hub-and-spoke conspiracy.

Since then, the Third Circuit reversed. I want to be careful about what that means. This is a private civil case. The decision concerns whether the allegations in a complaint were sufficient to survive a motion to dismiss. It does not establish what actually happened. The litigation continues, and I am not going to comment on the parties.  But the way the court analyzed the allegations is worth attention.

The Third Circuit focused on several allegations. Competitors allegedly provided current, nonpublic pricing and occupancy information to a common platform and understood that their competitors were doing the same. They then received recommendations generated using that collective information and allegedly followed those recommendations at a very high rate. Taken together with the other allegations, the court concluded that was sufficient at the pleading stage to support an inference of agreement.

The court had quite a bit to say about artificial intelligence and about the ways newer technology may present new factual questions. But when it came time to analyze agreement, it applied familiar Section 1 principles. And one part of the opinion makes that point especially well. The defendants argued, among other things, that the hotels remained free to override the software’s recommendations. The court’s answer relied on a principle from 1942: a price can be fixed by agreement even if every conspirator does not adhere to the agreed price every time.

That is an old rule applied to a new mechanism. Technology may change the facts we have to analyze. It does not change what counts as an agreement. And that remains the central point for criminal enforcement as well. In Cornish-Adebiyi, the question was whether the allegations made an agreement plausible. At a criminal trial, our burden is very different: we have to prove an agreement beyond a reasonable doubt.

The core question, though, is still the same: Did the competitors agree?

For companies using these tools, the practical advice is straightforward: know what data go in, where the outputs go, and whether non-public competitor information is being pooled or fed back into pricing decision. Bring antitrust compliance into the deployment of pricing and revenue-management tools, document that review as it happens, and train the people who use them on the same old rule: competitors still have to make independent competitive decisions. An AI governance process that covers privacy, cybersecurity, and other risks but never asks the antitrust question is not enough. 

Detection Tools

And if that compliance process identifies a potential problem, speed matters. That brings me to detection. For decades, the Antitrust Division’s Leniency Policy has been one of the central tools of cartel enforcement. It creates an extraordinarily powerful incentive for a participant in criminal conduct to identify the problem, come to us quickly, and cooperate.

And I want to emphasize quickly. Our published guidance tells companies to seek a marker at the first sign of potential wrongdoing — even before they know for certain that a violation occurred.

There is a reason for that. Leniency is deliberately designed to create a race to our door. Only one organization or individual can receive leniency for a conspiracy. We want companies that discover possible cartel conduct asking themselves who is going to get to the Antitrust Division first — not whether they can spend another month investigating before they have to make that decision.

The Whistleblower Rewards Program creates another race to the same door. Those are different races. A whistleblower reward is not a leniency marker. Someone with original information may instead have an independent financial incentive to come directly to us. 

But the races interact. Information reaching us through one channel can change what remains available through another. And that matters to a company that is still deciding what to do.

I view the two programs as complementary. Leniency gives a participant in a conspiracy a powerful reason to identify misconduct and move first. The Whistleblower Rewards Program gives someone who knows about the conduct an independent reason not to wait for the company to do so.

And that dynamic is particularly important when we start talking about conduct implemented through algorithms or other sophisticated technology. You will hear the argument that algorithms can make coordination quieter: fewer phone calls, fewer meetings, fewer traditional cartel communications.

Maybe. But quieter is not the same as invisible. In some respects, technology can disperse the evidence more widely. An arrangement implemented through an algorithmic pricing system may touch engineers, data scientists, product managers, sales personnel, vendors, compliance personnel, and executives.

No single one of those people necessarily knows the whole story. An engineer may understand the inputs without knowing whether any competitors agreed. A salesperson may know what customers were told without knowing how the system was built.

But pieces of the relevant evidence may be distributed across many more people than in the traditional image of a cartel conducted by a handful of executives behind a closed hotel-room door. That means more people may have the relevant pieces of evidence. The Whistleblower Rewards Program gives them a reason to bring those pieces to us. 

So, if you represent a company and you find any indication that your client may have been involved in criminal antitrust conduct, the message is not: finish your investigation and then decide whether to call us.

The message instead? Call us.

You do not need a finished investigation to seek a marker. Our published guidance deliberately sets a low threshold because the marker exists precisely for situations in which the company does not yet know the whole story.

The internal investigation can continue.  The clock will too.  And that only matters when there is a credible threat of enforcement.

Credible Threat of Enforcement

The Antitrust Division has now won four consecutive criminal jury trials. Three of those trials produced Sherman Act convictions. And in three of the four, the government presented its case-in-chief in five days or fewer. The most recent involved a five-year price-fixing conspiracy targeting over $100 million in publicly funded transportation construction contracts across Oklahoma.

I noticed that a law firm recently highlighted this record in a client alert. I do not usually rely on the defense bar to advertise the Division’s trial results for us, but I was happy to accept the assistance. The point, though, is not the streak. Trial records change. Every case depends on its own evidence. And anyone who has tried cases to juries knows better than to assume that yesterday’s result tells you what happens tomorrow.

And trials aren’t the only measure of successful enforcement. Our cases, like the rest of federal prosecutions, overwhelmingly resolve through plea agreements. Whether it’s price fixing in Philadelphia on public transit parts or procurement collusion on defense contracts in Hawaii, Maryland, and elsewhere, guilty pleas are part of the credible threat of enforcement. Those of you who have done this work, whether on behalf of the U.S. or paying clients, know all too well that the first plea agreement in an investigation is often the beginning, and not the end, of the enforcement story. 

Whether it’s a verdict after trial or a guilty plea, the important point is the same: we are willing and able to obtain convictions. Prosecution is not an empty threat. It is the consequence of waiting too long. That credible threat is what gives Leniency value. And a whistleblower’s information matters because it can lead to an investigation, and an investigation can lead to a case the government is prepared to prove to a jury.

And, to me, the length of those recent trials is almost as interesting as the results. Because criminal antitrust investigations can become extraordinarily complicated. Add AI or algorithmic conduct and you may have enormous datasets. Source code. Model architecture. System logs. Internal communications. Vendor communications. Experts. Technical witnesses.

There can be an enormous amount of complexity between the beginning of an investigation and the courtroom. But when we charge a case, our task at trial remains remarkably straightforward.

• Who agreed? To what?
• What did they know? What did they do?
• And can we prove it beyond a reasonable doubt?

Complexity is not an excuse for failing to tell a clear story. AI does not change that either. In May, I talked about what happens when the hub of a pricing arrangement is not simply conventional pricing software, but a large language model. My answer has not changed. The question is still whether competitors agreed.

Conclusion

If there is one theme connecting all of this — the Third Circuit’s decision, leniency, whistleblowers, and the Antitrust Division’s recent trial experience — it is this: The technology may be new.  It may change how the conduct works.  Where the evidence sits and who can see it. Even how quickly it reaches us.

But it does not change the fundamental legal question. The government still has to prove an agreement. And where competitors actually agree to replace independent competition with coordination, they do not get a different rule because the agreement traveled through a platform, an algorithm, or a large language model.

New code. But an old crime.

Thank you.