Doomsday, Fortnite And More Leaks Overnight

While the slow drumbeat of AI marches on, there are moments that are not slow, gradual progression, but giant leaps forward, for better or worse. Google’s new Nano Banana Pro image generation system, which launched this week, is that kind of change. And boy, is it wreaking havoc online.

There are, of course, larger questions about global disinformation, but in my smaller, less important corner of the world, what Google’s new image generation has done to the concept of movie, TV and video game leaks overnight is something I’ve never seen before. The entire concept of leaks is essentially dead.

For decades, there have always been “spy shot” glimpses at various projects, accidentally released imagery, video games putting things live too soon, or internal leaks circulating. If you’re faking it, it was a lot of work and often some questionable use of Photoshop. Now? Nano Banana Pro is so good that either leaks are indistinguishable from reality, or at the very least, they have discredited actual leaks because everyone believes they’re probably AI. Studios must be loving this.

Here are some of the most prominent examples that have racked up millions of views in the last 48 hours alone:

X-Men in Avengers Doomsday

Play Puzzles & Games on Forbes

The Boys season 5

Fortnite Chapter 7

There are plenty more and again, this is just the last few days. As you can see, these range from blurry leaked shots, to oopsie social media posts to what appear to be literal publicity stills from The Boys. I have had to parse AI and real images for over a year now as this tech grows, but even I was almost sold on those season 5 images. If I didn’t know Nano Banana Pro had just launched, I probably would have 100% assumed they were real. To see how easy this was, I tested it out myself and was able to generate a believable image of Hugh Jackman’s Wolverine filming on a Doomsday greenscreen in about four minutes with three prompts.

There is, of course, a dedicated collection of AI hunters who expose these, and some are worse quality than others, but I’ve also seen this group go after actual leaks that are real, and the ultimate message is simple: You just cannot trust anything now in either direction. Then, of course, we move into reality more generally, and Google has apparently put no guardrails on any public figure, unlike other models:

This reminds me of what has happened to art specifically the last couple years. Lots of AI art is easy to spot, but some is not, and it has badly damaged actual artists constantly accused of using AI to the point where they have to actually record themselves drawing to prove it’s legitimate. But we’re effectively at a point where even that can be faked now. Now, image generation has gotten truly photoreal and we are in an entirely new era that I don’t think people fully understand yet.

At least with the leaks situation, there are no real victims in the same sense. But to watch an entire corner of online media transform overnight because of a single new AI release has been something to behold. I can’t look at everything twice now. It’s gotta be five times, ten. Or we just give up entirely, which may be the only real option.

Follow me on Twitter, YouTube, Bluesky and Instagram.