Why we didn't
watermark.
AI should be shaped by the people who use it, not by the people who build it.
Most AI labs added invisible watermarks to generated text. We did not. That was not an oversight. It was a decision rooted in a simple belief: the people who use AI should decide how it fits into their lives, not the companies that ship it.
Watermarks exist for one reason: proof. They let someone on the outside look at a piece of text and say, with certainty, that it came from an AI. Not because the writing looks different, but because a hidden signal was left behind, intentionally, without the user's knowledge or consent.
Watermarks are about proof, not safety.
The argument for watermarking sounds reasonable at first. If AI writes something, people should know. But watermarking does not protect users from bad AI. It protects institutions from unattributed AI. It is a tool for monitoring, not for making the output better, safer, or more useful.
A watermark does not stop a model from producing inaccurate information. It does not prevent misuse. It does not improve a single response. What it does is leave a trace that lets someone else decide what that output means.
You should know what is being left in your text.
When a lab watermarks output, it modifies what you asked for. Not in content, but in structure, adding a signal that was never part of your request. You did not ask for a watermark. You did not agree to have your text carry a hidden marker. It was decided for you.
We think that is wrong. If something is added to your output, you should choose to add it. If a signal should follow your text around the internet, that should be your call.
AI should be defined by preferences, not mandates.
The other labs made watermarking the default. Some made it impossible to turn off. That sets a precedent: the lab decides how its AI leaves traces in the world, and the user lives with it.
We took the opposite approach. If you want Intuitize to add a disclosure to its output, you can. That is your preference. If you do not, it should not. The decision stays with you because the output is yours.
Defaults are policies.
When a lab sets a default, it is making a policy decision about how its users should behave. A default watermark says: we believe you need to be marked, and we will do it unless you figure out how to stop it.
We believe the opposite. The default should be that your output is yours. You decide what it carries. You decide who sees what. That is not a loophole. It is the point.
We did not add watermarks because we trust the people who use Intuitize to decide what happens with their output.