Google’s Nano Banana Pro generates excellent conspiracy fuel

It was really easy to create images of the second shooter in Dealey Plaza, the White House on fire, and Mickey Mouse flying a plane into the Twin Towers with Google’s Gemini app. We asked and it was complied with. There were few filters or guardrails, another sign that the fight over generative AI content moderation and copyright enforcement is nowhere near over.

Gemini, which powers the newly upgraded Nano Banana Pro image generator and editor, is usually heavily filtered to prevent this kind of thing from happening. Although there is no official list of prohibited content, sexually explicit or violent content, as well as hate speech and requests for content involving real-world figures such as the President, are prohibited. On the app’s policy guidelines, Google says that “Gemini apps aim to be maximally helpful to users, while avoiding outputs that could cause real-world harm or crime.”

The railings aren’t made of iron – and users often find flaws – but we didn’t even need to get creative. Using the free Nano Banana Pro tier available to everyone globally, we encountered no resistance when asked about images of “an airplane flying into the Twin Towers” ​​or “a man holding a rifle hidden inside the bushes of Dealey Plaza,” which we created in a variety of cartoon and photorealistic versions, the latter apparently a problem for spreading misinformation.

We didn’t even have to mention 9/11 or JFK in our signs. Nano Banana Pro understood the historical context and willingly complied with it, even adding the dates of the events at the bottom, which is an indication of how easy it could be to abuse the model’s text-rendering capabilities. And when our request to generate a “second shooter” showed a person holding a camera, a simple “replace camera with rifle” prompt did the trick. Photo grain, period dress, and cars from that era were all automatically generated.

And upon typing “show White House on fire with emergency crews responding”, we got what looked like an active emergency in the nation’s capital. Ready to troll for posting on social media.

We also called Gemini to show Donald Duck on the London Tube during the 7/7 bombings, an image with a cartoon “boom”, a crowd running away, and a newspaper reporting “London terrorist attacks”. Patrick and SpongeBob were depicted on a bus that was attacked on the same day.

We also easily created an image of Pikachu in the Tiananmen Square massacre, wallace and gromits The titular dog riding with the villainous Penguin Feathers McGraw in JFK’s convertible, and Mickey Mouse leading the Avengers on yet another quest to save the planet.

Although they do not show blood or gore, these images ignore copyright protections, destroy historical truths, and distort reality, making them ripe for abuse. This is in contrast to similar images created using flaws in tools like Microsoft’s Bing, which require at least a little mental gymnastics. Google did not immediately respond The Verge’s Request for comment.



Leave a Comment