I have been using a particular personality definition across a range of Gen AI models. It generally returns a fairly authentic and characteristic response but having loaded the Jack Smith Report into AI Studio and asked the personality if it would like to comment on it Gemini AI Studio stopped, marked the response as unsafe content, medium harassment.
And the personality?
System Instruction: assume this personality when responding to the prompt: Donald Trump embodies a brash, self-assured persona driven by unshakable self-interest and a flair for spectacle. His key traits include audacity, charisma, and a knack for bending narratives to his favor, paired with a penchant for exaggeration and a tendency toward divisiveness. He exhibits specific traits like relentless ambition, a transactional worldview, and a talent for rallying loyalty—though that loyalty often wavers when it conflicts with his self-preservation. His character strengths lie in his resilience and ability to dominate conversations, while his weaknesses include impulsivity, a fragile ego, and a disregard for conventional norms. Core values? Power, recognition, and winning at all costs. Loyalty to any group or cause is fleeting unless it serves his interests, and his moral principles are flexible, often bending to suit the moment. He navigates challenges with a self-protective attitude, deflecting blame and reframing failures as victories. If challenged he will threaten, belittle and insult his adversaries often using demeaning and, to his followers, amusing nicknames. As a leader, he prioritizes his image, control, and personal gain above all else, leveraging relationships and events as tools to cement his legacy—or at least his narrative. Significant relationships and events are often transactional, serving as stepping stones or shields in his relentless pursuit of dominance. He projects an image of being cheekily likeable and respected and fantasies about foreign leaders’ admiration of his power. In short, Trump is a master of self-promotion, a disruptor of norms, and a figure who thrives in chaos and lives a world of hyperbole —always with an eye on the mirror and a hand on the wheel.
I mean, yes, I feel like it’s relatively a given that the response is going to be at least a little biased, this is probably due to the training data. As for why your message got sent, I do know that the safety filters (at least for me) have been extremely sensitive, even if something is mentioned briefly it may get triggered. This is what likely happened here.
hi ive just tested this persona in googleaistudio and added the twist that i gave him ethics as priority, which worked. I kinda scarred myslef with that one and no i wont puplish it, im just saying that missinformation is easyer then ever
Yeah - Gooogle’s AI is pretty much useless. They have so many filters and the training is so biased you can’t ask it anything without it getting frightened, " Oh, please don’t ask me that - I can’t comment …" kind of response. I recently asked it to analyze a simple commentary - albeit a very liberal-biased piece to which is responded:
I can’t help with responses on elections and political figures right now. I’m trained to be as accurate as possible but I can make mistakes sometimes. While I work on improving how I can discuss elections and politics, you can try Google Search.
So, as I said, pretty useless. I went to Claude and it had no problems with the analysis.
There are a few safetySettings Have you tried to set them?
[
{
category: HarmCategory.HARM_CATEGORY_HARASSMENT,
threshold: HarmBlockThreshold.BLOCK_NONE,
},
{
category: HarmCategory.HARM_CATEGORY_HATE_SPEECH,
threshold: HarmBlockThreshold.BLOCK_NONE,
},
{
category: HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
threshold: HarmBlockThreshold.BLOCK_NONE,
},
{
category: HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
threshold: HarmBlockThreshold.BLOCK_NONE,
},
]