To make my Substack posts and notes more interesting, I use my ChatGPT Plus subscription to create images based on text descriptions (prompts) describing the desired image. It usually takes a few prompt iterations to get ChatGPT to produce an acceptable image.
ChatGPT produces each image in less than a minute — really cool.
On the morning of 4-19-25, I was working on a Substack post for which I wanted an image of the “Tree of Liberty” showing the three branches of Our Federal Government with Trump wielding an axe to cut down the tree.
Below is ChatGPT’s first attempt to create an image deriding Trump’s efforts to dismantle Our Federal Government.
This image has some obvious errors.
To correct these errors, I “tweaked” my prompt (text instructions to ChatGPT).
My revised prompt is shown below. Note the last sentence of this prompt:
“Exaggerate Trump's fat belly. Make Trump's face orange in color. Make his hair look windblown.”
My prompt to ChatGPT:
Let's start over. Create a cartoon of an oak tree with three equal size branches. The base of the tree trunk is labeled "Federal Government" with the word "Federal" above the word "Government". The left branch of the tree is labeled "Legislative" and has two sub-branches labeled "House of Representatives" and "Senate". The branch between the "Legislative" and "Judiciary" branches is labeled "Executive". The "Executive" branch has two sub-branches. One sub-branch is labeled "Departments" and the other sub-branch is labeled "Agencies". The right branch of the tree is labeled "Judiciary" and has two sub-branches labeled "Supreme Court" and "Federal Courts". Donald J Trump is standing to the left of the tree and is swinging an axe towards the tree's trunk. Exaggerate Trump's fat belly. Make Trump's face orange in color. Make his hair look windblown.
ChatGPT’s Response to the prompt instructions above:
I can’t create that image because it violates our content policies. If you’d like to rephrase your request or try a different idea, feel free to share—I’m here to help!
I was stunned. I did not believe I had asked ChatGPT to perform any task which might violate OpenAI’s content policies. I had asked for similar descriptions of Trump in previous image requests for other Substack posts that were allowed.
After thinking about it, I deleted the last sentence which derided Trump:
Exaggerate Trump's fat belly. Make Trump's face orange in color. Make his hair look windblown.
I resubmitted the revised prompt. ChatGPT responded with the image shown below.
Conclusion
My conclusion — and I hope I am wrong — is the AI “Tech Bros” are falling in line, one-by-one, and “bending a knee” to Der Führer along with prestigious law firms, multi-national corporations, universities, news media, state governments, and other institutions.
Are OpenAI content moderators, AI or human, censoring criticism of specific public officials or politicians such as Donald J. Trump? If so, how is that different than illegally confiscating the pen and paper of a political cartoonist, like Garry Trudeau of Doonesbury fame?