A Grok summary of Mike Gundy's comments that attributed a false quote to him. A Grok summary of Mike Gundy’s comments that attributed a false quote to him. (X/Twitter, Candace Ward/USA Today Sports.)

There have been a lot of questionable results from generative artificial intelligence tools to date.

Some of those have come from X/Twitter’s “Grok” AI-powered summaries of news stories. And some of those have been in sports, including a complete missing-the-point with “LeBron Accused Of Affair With Teammate’s Mom” (following jokes about the Lakers drafting LeBron’s son Bronny).

But the latest example is the complete manufacturing of a quote (one that could perhaps be legally actionable). That comes from Grok’s summary of the current controversy around Oklahoma State Cowboys football coach Mike Gundy. For a good portion of Wednesday, the headline of X’s summary attributed a quote to Gundy that he never uttered.

Grok's summary of Mike Gundy's comments on X
Credit: X/Grok

One area where AI has thus far proven far worse than humans is in evaluating nuanced distinctions. Those are critical in the discussion around Gundy.

This comes from what Gundy said at the Big 12’s Football Media Days Tuesday in defense of not suspending running back Ollie Gordon II after Gordon was charged with driving under the influence.

“I looked it up on my phone- what would be the legal limit? In Oklahoma, it’s 0.08. And Ollie was 0.1. So I looked it up and it was based on body weight. Not to get into the legal side of it, but I thought, really, two or three beers, or four. I’m not justifying what Ollie did. I’m telling you what decision I made. Well, I thought, I’ve probably done that a thousand times in my life. And, you know, it was just fine. So I got lucky. People get lucky. Ollie made a decision that he wished he could have done better. …Is suspending him for one game really going to matter? I don’t think so.”

Gundy then later tried to clarify on X that this “wasn’t a reference to something specific”:

Some people concluded from Gundy’s comments that he was admitting to past driving with a blood-alcohol content of more than the legal limit. But Gundy did not say “I’ve driven drunk.” And that’s the kind of misquote that could possibly pose a legal problem for X. (Consider that Brett Favre is currently trying to drag Shannon Sharpe back into court in a dispute over whether Sharpe saying he “stole money from people who really needed that money” was allowable hyperbole or actually defamatory.)

An editorial note that reads “This story is a summary of posts on X and may evolve over time. Grok can make mistakes, verify its outputs” may or may not be enough of a defense there.

At some point during Wednesday afternoon, the headline changed to read “Gundy’s DUI Defense: No Suspension for Gordon.” It’s unclear if that was a manual decision or if the AI changed it on its own.

This comes amidst larger discussions around generative AI, and whether current versions of it actually have value worth the amount of computer and power resources being spent on it. Goldman Sachs analyst Jim Covello wrote earlier this week he doesn’t think they do.

With X at least, the Grok AI summaries don’t seem to be providing a lot of value. And they could instead provide some legal liability. (That has led to positive X/Twitter changes in the past, with a 2009 lawsuit from former MLB manager Tony La Russa prompting the creation of verified accounts.) We’ll see if anything comes of this one.


About Andrew Bucholtz

Andrew Bucholtz has been covering sports media for Awful Announcing since 2012. He is also a staff writer for The Comeback. His previous work includes time at Yahoo! Sports Canada and Black Press.