If you’ve spent even five minutes on social media lately, you’ve probably seen it.
People asking AI to “create a caricature of me based on what you know about me.”
The result is usually clever, flattering and oddly accurate. A digital mirror that reflects your personality, habits and style back at you as an avatar or character.
It’s fun. It’s shareable. It’s a little addictive.
But beneath the novelty sits a much more serious question.
What are you actually handing over when you ask AI to define you?
Before you jump on the trend, here’s a grounded reality check.
Many people take character creation a step further by uploading photos to “help the AI get it right”.
That’s where things shift.
Your face isn’t just an image. It’s biometric data.
When you upload high-resolution photos, you are effectively providing a facial recognition blueprint. On many free platforms, those images may be stored, reused or used to train future models. Even if the platform says it’s anonymised, the risk doesn’t disappear.
Unlike a password, your face can’t be reset.
Once it’s out there, it’s out there.
To get a deeper or more accurate character, people often share personal details.
First pets.
Childhood streets.
Family nicknames.
Where they grew up.
On the surface, it feels harmless.
In reality, these are the exact answers used in account recovery for banks, emails and government services. Feeding them into an AI conversation builds a profile that could be exploited through social engineering or identity theft.
You’re not just telling a story.
You’re filling in security forms without realising it.
A common assumption is, “I didn’t give it anything specific.”
AI doesn’t need specifics. It needs patterns.
A job title here.
A general location there.
A few hobbies.
A tone of voice.
Individually, they mean nothing. Together, they form a surprisingly accurate picture.
This is known as the mosaic effect. Small, vague details combine to identify a real person. At that point, you’re no longer a fictional character. You’re a traceable data point.
There’s an old saying in tech that still holds true.
If you’re not paying for the product, your data is the product.
Free AI tools often use your prompts, conversations and images to train future models. That means the quirks you thought were private can become part of a broader dataset.
Paid platforms are not perfect, but they usually offer stronger privacy controls, clearer opt-out options and temporary or non-training modes.
The difference matters more than most people realise.
You don’t need to avoid AI character creation altogether. You just need to be intentional.
Use persona, not person
Ask for a character based on a vibe or archetype rather than your real life.
Think “a futuristic consultant who loves espresso” instead of your actual background.
Mask your data
Never use real names, real pet names, or specific locations. Swap them out or keep them generic.
Check your controls
Turn off chat training where possible. Use temporary chats if the platform offers them. These settings exist for a reason.
AI caricatures are entertaining, creative and here to stay.
But every prompt you write is still data.
Before you ask AI to define you, pause and ask yourself a better question.
Would I hand this information to a stranger?
Because in the digital world, that’s often exactly what you’re doing.