We are in a whole new era of movie and TV “leaks” thanks to the advancement of AI, namely because of a new tool, Google’s Nano Banana Pro, the just-launched image generator that can replicate real life almost exactly, and can do so using the likeness of infinite celebrities, 1:1.
Now, this has already gotten so widespread that heads of entire comic book universes are forced to comment. In this case, a pair of images became widely circulated online, showing the stars of the upcoming DCU HBO Green Lantern show, Lanterns, Aaron Pierre and Kyle Chandler with power rings. Eventually, James Gunn himself had to chime in to say that nope, these were in fact AI.
It’s just one image, but the internet is being flooded with thousands of these, particularly in the entertainment sector, claiming early looks at future Marvel and DC projects that are usually no longer in the “AI slop” territory, but genuinely difficult to separate from reality.
The problem is not just that these are “good” fakes, it’s that they’ve become so easy to make. With Nano Banana Pro, all you need for something like this is to upload the only official image from the show (seen at the top of this post) and tell the AI to use it as a reference showing these actors using power rings in a similar setting.
I won’t post them here, but in a few minutes, I made a few images that are about 95% of the way there to the Kyle Chandler shot. Previously, testing out this tech, I’ve made convincing shots of Gunn behind the scenes of a Wonder Woman movie with the character seen in shadow. I’ve made a believable costume fitting for Batman’s cowl with actor Jacob Elordi. I could make a dozen of these in a couple hours, and I’m one person with limited experience with these tools (I’d argue mine are better than the Lanterns one here). You can imagine where things are headed now.
One big question mark here is why Google is allowing perfect 1:1 generations of celebrities and famous characters like superheroes, something that many other image and video generators have steered clear of, if not initially, then soon after they received complaints. This is the least amount of guardrails I have ever seen on an AI generation system, and I have to wonder if Google is going to be forced to change things, which would disrupt this fake leaks pipeline, not to mention wider disinformation generation across politics and such.
We do have some legitimate info on Green Lanterns taken from a recent HBO preview event. We know:
- Hal Jordan has been training John Stewart for a while as a new Green Lantern. John believes he’s better at saving people than Hal, which he does not appreciate. Both Chandler and Pierre’s performances were well-received.
- Despite being “grounded,” there are traditional Green Lantern moments with ring generations and flying. Mostly, the cinematography is very True Detective-inspired.
- The character Ch’p is mentioned but not shown.
Lanterns will air later next year after Supergirl in the summer, but expect many more attempts at “leaks” like this. It may force studios to release legitimate images to curb confusion.
Follow me on Twitter, YouTube, Bluesky and Instagram.
Pick up my sci-fi novels the Herokiller series and The Earthborn Trilogy.
Source: https://www.forbes.com/sites/paultassi/2025/11/26/dcu-green-lantern-leak-that-fooled-fans-shot-down-by-james-gunn-as-ai/
