Twitter’s AI Creates Deep Fakes of Politicians and Even Elon

Twitter’s AI tool, Grok, has very limited “guard rails”. For example, one Grok AI image shows Trump piloting a plane that looks like it is about to crash into the Twin Towers with one of the towers already in flames.

Even though Grok says that it won’t let you create images that infringe on copyrights and trademarks, it is perfectly willing to create Disney or Nintendo characters, in any quantity. It is not clear how copyright law would apply in this case. Twitter would say that they provide the tool and the users are infringing on the copyright but Disney might say that they are facilitating copyright infringement, which is a crime in and of itself.

Even Elon is not immune. Pictures of Elon hold a sign that says “gender affirming care is lifesaving” have been posted on Twitter. Elon has a trans child and he has disowned her and she doesn’t want to have anything to do with him.

Congress is asking Twitter and Meta to talk to them about AI generated deep fakes of politicians ahead of the elections. They are not and the result may be laws that they don’t like very much.

Examples of non-political deep fakes include the Pope in a puffer jacket and Tom Hanks hawking dental insurance.

On the other side of the pond, deep fake videos of UK Prime Minister Keir Starmer and Prince William are showing up on Meta. They are hawking fraudulent crypto schemes. For example one ad says:

“Good afternoon, honored citizens of the United Kingdom,” stated the deepfake Prince William. “I am pleased to announce that I, Prince William, and the entire Royal Family fully support Prime Minister Keir Starmer’s initiative and his new platform.”

In the absence of any laws, you should assume that we will see a lot more of this the closer we get to November.

Credit: Mezha and Dark Reading

by