Mark Wahlberg Wants a Backup Copy, With Family Holding the Keys

Mark Wahlberg’s rule for AI in movies fits in one line. Use it “to enhance what you’re doing.” He said it on stage at TechCrunch Disrupt on October 16, 2026, in conversation with TechCrunch Editor-in-Chief Connie Loizos. Then he went further than most actors will go in public. He said he’s open to an AI version of himself carrying on his legacy. He also didn’t object to the idea of a digital Wahlberg keeping the Ted series going for decades.

I spend most weeks testing AI tools and writing down exactly where they break. So my first reaction was that Wahlberg had just described a product spec. It’s a loose one, but it’s a spec. “Enhance what you’re doing” is the same standard I hold every writing assistant, video generator and voice model to. Does it make the work better, or does it only make more of it?

Enhancement is a testable claim

On this site, the tools that earn a recommendation are the ones that fit into a workflow someone already has. They speed up the boring parts and stay out of the way during the parts that need a human. The tools that fail usually try to replace the person instead of helping them. They produce something that looks finished from across the room and falls apart up close.

Wahlberg’s framing lines up with that split. It’s also the easiest promise in tech to make and the hardest to keep. Every AI company says its product enhances creators. The proof is in what happens on a real project, under a real deadline, when the output has to hold up in front of an audience. Hollywood remains divided over AI, and from where I sit that division makes sense. “Enhance” means very different things depending on whether you’re the one using the tool or the one it might replace.

Wahlberg wears several hats here. He’s an actor, a director and an angel investor. That gives him a view of AI from the talent side, the production side and the money side at once. It also means his optimism deserves the same scrutiny I’d give any vendor pitch. That isn’t cynicism. It’s the job.

A likeness with a lockbox

The more interesting part of the conversation was about legacy. Wahlberg pointed to the CAA Vault as a possible route for an AI version of himself, with his family in control of his likeness. In toolkit terms, that’s the access-control layer. In my experience, that layer decides whether a product is trustworthy far more often than the model behind it does.

I haven’t had hands-on time with the CAA Vault, so I can’t say how it works under the hood. I can say what I’d want answered before I’d recommend any likeness-storage product, to a movie star or anyone else:

• Who holds the keys, and what happens when the people holding them disagree?
• Is there an approval step for every use, or a blanket license once the likeness is stored?
• Can permission be revoked, and does revoking it actually stop existing outputs from circulating?
• Is there a clear record of where and how the likeness has been used?
• Who judges whether a given performance “enhances” the legacy or cheapens it?

None of those are model questions. They’re governance questions. In my experience, governance is where AI products quietly succeed or fail. A tool can generate a flawless digital double and still be a bad product if nobody can explain who’s allowed to press the button.

Putting family control at the center is the right instinct. It treats a likeness like property with heirs instead of a dataset with a license. That framing alone is more thoughtful than a lot of what I see in consumer AI, where your face or voice often becomes training material through a checkbox you never read.

Ted for decades is a stress test

There’s some irony in which franchise came up. Of all the films an AI Wahlberg could carry forward, the one discussed was a series about a talking teddy bear. But “for decades” is the phrase that caught my reviewer’s eye.

Long-running consistency is where the AI video and voice tools I test still struggle most. A clip can look great. A whole project that stays believable from start to finish is much harder. Keeping a character recognizably the same across years of new material is a different problem. It needs taste, continuity and someone accountable for every creative call. A franchise running for decades on a synthetic lead would be the hardest stress test any likeness system could face.

That loops back to Wahlberg’s own rule. If the AI version is there to enhance what a creative team is doing, decades of Ted might be workable. If it becomes a way to keep a franchise running without anyone actually steering it, it fails the standard he set.

My verdict

What works is the framing. “Enhance, don’t replace” is the right test, and family-controlled likeness puts consent and ownership where they belong. What doesn’t work yet is the detail. A stage conversation can’t answer the governance questions that decide whether this is a solid system or an expensive promise. Hollywood’s divide over AI won’t close because one star is relaxed about it.

Still, I’d rather review a product whose maker states the standard out loud. Wahlberg did. Now the tools have to meet it.

🕒 Published: October 18, 2026