DISCOVERY MELTDOWN AT AMEX: Duke Decision Changes the Game

Here is a follow-up on one of the biggest ongoing TCPA cases we have been watching this year. Back on July 11, 2025, the District of Arizona issued a discovery order in Debra Duke v. American Express Company that sent shockwaves through the compliance world. It forced AmEx to turn over call data tied to the wrong party and never call numbers and instantly became the ruling plaintiffs’ hold up when arguing that outbound call logs must be produced in putative TCPA class actions.

To jog your memory, Duke is the case where the plaintiff says American Express repeatedly blasted her with prerecorded calls that were meant for someone else entirely. She was not a customer. She told them that. The calls continued anyway. So she sued on behalf of a putative class and then sought the core data every TCPA plaintiff wants. She asked for call logs showing how many times AmEx dialed the wrong party or never call numbers with an artificial or prerecorded voice. AmEx pushed back and insisted its systems could not readily determine how many such calls occurred and that the requested discovery was overbroad and unduly burdensome. The Court did not accept that and ordered production, finding the information central to Rule 23 issues like numerosity and commonality.

That brings us to the latest development, which turns the pressure up even higher for American Express and for every large caller facing wrong number exposure.

American Express just took a real hit in this follow up order and it should make every major caller a little uneasy. The newest ruling essentially confirms that the Court is not giving AmEx a shortcut, not giving it the benefit of any technical limitation, and not allowing it to shield call data that it claims is too difficult or expensive to retrieve. The Court has made clear that if you are making high volume outbound calls with prerecorded or artificial voice technology, you should expect to be able to account for every one of those calls, including calls made before a number was flagged as incorrect.

The heart of the problem for AmEx is that its own corporate representative testified that the company could in fact run searches for calls that resulted in wrong number notations and could identify which of those numbers received prerecorded calls. That testimony undercut the narrative that AmEx had been trying to maintain. The Court saw the gap between the testimony and the company’s objections and determined that the discovery AmEx had already produced was too narrow, was not generated using the process the Court had expected and did not satisfy what was required under the earlier order.

Plaintiff took that discrepancy straight back to the Court and filed a second motion to compel. AmEx again claimed burden and again argued privacy and proportionality. It even filed a declaration estimating that compliance would require searching roughly 100 terabytes of data, taking sixteen weeks and about 100,000 dollars of internal resources. The Court did not accept that position either. It contrasted the elaborate burden described in the declaration with the more efficient process the company’s own witness said was available. The Court found that AmEx had not shown the requested discovery to be unduly burdensome and that the information was essential for class certification.

The privacy argument did not gain traction either. The Court pointed out that plaintiff was not asking for account details or financial information. She was asking for phone numbers and call counts, and a protective order was already in place. In this TCPA context, that was not enough to block production.

So, the Court granted the second motion to compel in full and ordered American Express to produce the missing discovery by August 15, 2025. It also extended the class certification schedule to make sure plaintiff has time to analyze the data before briefing. That is a significant signal because it shows the Court believes this information is genuinely necessary to evaluate whether a class can be certified. It is not a side issue. It is not a fishing expedition. It is central to the case.

For the industry, the takeaway is clear. If you dial wrong numbers with prerecorded or artificial voice technology, even unintentionally, the call logs connected to those numbers are fair game in discovery. And courts are willing to scrutinize your technical explanations and compare them to what your own witnesses say. If there is any inconsistency, you will be ordered to dig deeper.

From a risk standpoint, Duke shows why wrong number prerecorded calls can quickly transform from a one plaintiff dispute into a major class action threat. Once a plaintiff can access a list of wrong party or never call numbers and match them against outbound call data, the potential exposure grows immediately. A defendant is no longer arguing about one consumer’s experience. The defendant is staring at the possibility of thousands of calls at 500 to 1,500 dollars each.

Bottom line, the Duke follow-up confirms that discovery obligations in TCPA cases are not limited by the fact that a company’s data systems are complicated or expensive to query. If you choose to place large scale prerecorded calls, you are expected to maintain the kind of data that will allow you to demonstrate what you did and to whom. And if you cannot do that easily, courts are increasingly willing to treat that as your problem, not the plaintiff’s.