Elon Musk claims people manipulated Grok into paying him delusional compliments with ‘adversarial prompting’ – 17 straight-talking fact-checks
The past few days have seen Elon Musk’s AI bot Grok make some outrageous claims about the controversial tech billionaire’s physique, intelligence, and skills.
Grok says Elon would have created the cosmos more quickly and efficiently than God in the Bible. pic.twitter.com/zOQ9448A3s
— Sturgeon's Law (@Sturgeons_Law) November 20, 2025
I can’t fucking breathe https://t.co/h1q8zpFymW pic.twitter.com/GjsdHHjpOS
— ɹoʇɐʇoɹ pɹoʍ (@wrotator) November 20, 2025
Grok thinks Musk should've been the #1 pick in the 1998 NFL draft. https://t.co/nCkZiws1Hc pic.twitter.com/qv0OIePc8C
— Jacob Silverman (@SilvermanJacob) November 20, 2025
In awe at how totally Elon has broken Grok. The man is a slave to the perception of others. https://t.co/iNOo8zXRnB pic.twitter.com/1gwTO0ue6z
— Hunter (@StatisticUrban) November 20, 2025
for posterity in case they go on a deletion spree like in previous times they've updated grok to be embarrassing pic.twitter.com/aNCUKOSZtk
— Colin Fraser (@colin_fraser) November 20, 2025
As predicted, they did indeed start deleting Grok’s mortifying sycophantic comments, and Musk posted his hot take on the whole embarrassing affair.
CW: It contains an ableist slur, disingenuously levelled against himself.
Is the adversarial prompting in the room with us now? Tweeters thought not.
1.
Let the history books show that I set in motion the chain of events that led Elon Musk to describe himself (correctly) as a fat retard. https://t.co/O6Mau8L59u pic.twitter.com/N09VFAsvLp
— ɹoʇɐʇoɹ pɹoʍ (@wrotator) November 20, 2025
2.
Partakes in bullshit and then acts like he was hacked when caught, I just know this guy was sweating https://t.co/fItjwBlZsx
— Souvik (@Souvik_art1) November 20, 2025
3.
People are literally just asking the chatbot questions. How is that it being "manipulated"?
— Wild Geerters (@steinkobbe) November 20, 2025
4.
Whenever you think this idiot couldn’t get more insecure he finds a way https://t.co/JbTtkRu5A0 pic.twitter.com/7GIenrtsIR
— Rahul "we can all track N628TS, Elon" Desai (@rahrdes) November 20, 2025
5.
He programmed Grok to support him no matter what and it immediately blew up in his face pic.twitter.com/zWGFDfC5VU
— evan loves worf (@esjesjesj) November 20, 2025
6.
"Adversarial prompting" is when I ask the exact same question twice but mention Elon Musk in one of the questions. https://t.co/yNPvPc3Ac7 pic.twitter.com/YKD5EABsUQ
— steven monacelli (@stevanzetti) November 21, 2025
7.
Where is the adversarial prompting? pic.twitter.com/L7gGRVy3gp
— Sheel Mohnot (@pitdesi) November 20, 2025
8.
Elon is now claiming people tricked grok into saying absurd things with adversarial prompting and has started deleting grok posts. Here's the "adversarial prompting" in question. If you ask any other chatbot these questions they answer LeBron James is more fit than Elon pic.twitter.com/jdzEMnCQ8e
— Wild Geerters (@steinkobbe) November 20, 2025
9.
“Adversarial prompting” lmaoo wut? How is asking “who is smarter” or “who is stronger” adversarial my guy. https://t.co/O5PVCjzF6e
— JP (@jpchen__) November 20, 2025