Deepfake videos on TikTok can be fun. They can also be malicious.

Facebook, Twitter, YouTube and TikTok have banned deepfakes, which are realistic but fabricated or manipulated representations, often of public figures. Yet a TikTok video in which a fake Tom Cruise serenades the real Paris Hilton went viral. That video is one of dozens from the account @DeepTomCruise.

The account has racked up almost 4 million followers with its digital simulations of the famous actor singing, golfing and, of course, laughing in a slightly too intense way.

Marketplace’s Meghan McCarty Carino speaks with Anjana Susarla, professor of responsible artificial intelligence at Michigan State University’s Broad College of Business. She said we can thank TikTok’s algorithm for keeping deepfakes alive. The following is an edited transcript of their conversation.

Anjana Susarla: Because we all love celebrity content like deepfake Tom Cruise and deepfake Keanu Reeves and stuff like that. And I think there’s some amount of latitude if a person is posting a video which is altered Tom Cruise, but it’s clearly for satire, then the creator may have a bit of a leeway around. But then the question is, what if people are using this with malicious intent?

Meghan McCarty Carino: Are algorithms or the other ways that platforms enforce their bans of deepfakes — can they make a distinction between different kinds of content: comedy versus what are some of the very obvious malicious uses of deepfake?

Susarla: No, I would say that you will actually need a lot of human intelligence. Has TikTok invested in these mechanisms where there are also people in addition to algorithms curating all these news feeds that the users see? We don’t have any answers. They haven’t put in place enough content moderation policies, or at least they have not been transparent. So we as the users have very little idea what they are doing.

McCarty Carino: And what about when deepfakes are being used, maybe even in comedic ways, but used in commercial advertising, as we’ve seen increasing?

Susarla: Yes. And that’s also a danger because what if the rivals can use deepfakes of their competitors? What is the position that courts will take in enforcing these provisions? I think those are all open questions.

McCarty Carino: Do you think we have the legal tools that we need at this time to contain some of the harms that all of this could cause?

Susarla: So let’s take this example of revenge pornography. It’s still not understood well enough. So deepfakes, coupled with things like that, would be very, very, very harmful. And I think that legal infrastructure maybe has not really — we have not really had enough awareness of all those harms that can spread.

Related links: More insight from Meghan McCarty Carino

The Wall Street Journal recently reported on the spate of advertisements featuring deepfakes, including the appearance of a fake Bruce Willis defusing a bomb in a Russian phone commercial and a very fake Elon Musk explaining a real estate investment startup in the style of Margot Robbie in the bubble bath scene from “The Big Short.” Honestly, it is as weird and creepy as it sounds.

As the article explains, public figures have a pretty clear-cut legal case when it comes to unauthorized use of their likeness, but the sheer ease with which this content can be generated could create what one expert described as death by a thousand cuts. Lawsuits just won’t be able to keep up with every instance of a deepfake ad.

Susarla raised concerns about how deepfakes also could be used in revenge porn. Last year, MIT Technology Review reported that while almost all states prohibit the distribution of nonconsensual pornographic material, only two — California and Virginia — include the use of deepfakes in those bans.