AI Unlocked My Voice. Watermarking Could Take It Away.

I was reading my email this morning when I came across a piece by Taylor Arndt. The title caught my attention: “AI Finally Gave Some of Us a Level Playing Field. I Don’t Want Watermarking to Take That Away.”

I read it. Then I read it again.

Taylor laid out the policy. The EU AI Act. Anthropic’s worldwide text watermarking. Google’s SynthID. The question of what happens when AI is used for OCR, for dictation, for helping someone with a cognitive disability communicate their own thoughts more clearly.

And then I hit the line that stopped me cold.

“What if somebody with a cognitive disability gives AI their own ideas and asks it to communicate those ideas more clearly?”

That’s me. That’s exactly me.

AI Isn’t One Tool. It’s Everything #

I’m blind. I have rheumatoid arthritis, idiopathic intracranial hypertension, fibromyalgia, and a neurological condition that affects my ability to organize my thoughts into words other people can follow. I live in a nursing home. I use a computer, not a phone, because touch screens don’t work well for me.

And AI is in nearly everything I do.

I write. A lot. I’ve published over 30 articles on my website about disability, technology, faith, and advocacy. I’ve submitted pieces to paying publications. Writing is my career path. The thing I found after 34 years of not knowing what I could do.

But I can’t write the way most people write. My thoughts are clear. Getting them into linear text that other people can follow is not. AI bridges that gap. I give it my ideas, my experiences, my voice. It helps me organize them into something readable. The thoughts are mine. The stories are mine. The words are a collaboration between me and a tool that makes communication possible.

That’s just the writing.

I run Hermes Agent on my computer because I needed an AI that could help with everything. It helps me plan my day when executive function fails. It helps me manage projects when my brain won’t hold the pieces together. It helps me communicate when words are hard. I didn’t set it up because AI is trendy. I set it up because I needed infrastructure to function.

I build things. I developed Everrealm, a screen-reader-first browser strategy game. I built WordPredictor and windowState, NVDA add-ons that help blind users type more efficiently and manage windows. None of that happens without AI-assisted development. I direct. I design. I test. AI helps me implement.

I maintain my website. I manage communities. I use TextExpander, which has AI built in, to reduce the physical and cognitive load of typing. I use Raindrop.io, which also has AI, to manage bookmarks across projects. Both connect to Hermes through MCP servers, so my AI assistant can work with the tools I use every day.

I use IFTTT to connect services across my life. It automatically sends new articles to my disability communities, making sure people who need the information can find it. I use Pushover to get my attention with notifications I can’t miss, because when you have executive function challenges, “I’ll remember to check” isn’t a strategy.

Hermes connects to Slack, where I’m building MultAbilities. A learning community for people with multiple disabilities. AI helps me plan, organize, and communicate with collaborators. It’s not replacing community. It’s making community possible.

On my phone, I use Seeing AI and other tools to get information about things I can’t see. A document. A label. A room. AI describes what my eyes can’t.

This isn’t a list of conveniences. It’s infrastructure. AI isn’t one tool I use for one task. It’s threaded through my entire life. Writing, coding, planning, communicating, building, managing, connecting. Without it, most of what I do wouldn’t be possible.

The Problem With Watermarking #

Anthropic is putting machine-readable watermarks into text generated by Claude. Google already does this with SynthID. The EU AI Act requires it. The direction is clear: more detection, more provenance, more ways to know when AI was involved.

I understand why. Deepfakes are real. Fraud is real. Impersonation is real. I’m not arguing that provenance technology shouldn’t exist.

But a watermark can only tell you one thing: AI processed this text.

It can’t tell you why.

It can’t tell you that the person who used AI has a neurological condition that makes organizing words difficult. It can’t tell you that someone dictated their own thoughts and used AI to clean up the transcription. It can’t tell you that a blind person used a vision model to read an inaccessible document. It can’t tell you that AI didn’t replace the human. It enabled the human.

But humans will interpret “AI detected” as “this person didn’t do the work.”

Schools will flag it. Employers will filter it. Publishers will reject it. Platforms will label it. Some future automated system will decide that content touched by AI is less trustworthy, less authentic, less valuable.

And disabled people who depend on AI to communicate will be the ones who pay.

What I Stand to Lose #

I found something I can do.

For 34 years, I didn’t know what that was. I tried coding. I tried traditional education. I tried paths that other people could walk but I couldn’t. My body and brain don’t work the way most systems expect them to.

Then I found writing. AI-assisted writing. The kind where my thoughts, my experiences, my voice are the source. And AI helps me get them into a form other people can read.

I’ve built a career on this. I’ve raised awareness about disability, about long-term care, about what it’s like to live in a nursing home at 34. I’ve built communities. I’ve developed games and tools. I’ve submitted to paying publications. I’ve found a way to contribute that works with my brain instead of against it.

And now there’s a movement to mark everything that passes through AI as potentially suspect.

Not to distinguish between “AI wrote this” and “AI helped a disabled person express their own thoughts.” Just to flag it. And once it’s flagged, the judgment begins.

If my articles carry a watermark, do they become less credible? If my game code carries a watermark, is it less legitimate? If my community messages carry a watermark, am I less present?

I don’t know the answer. But I know that the systems being built right now aren’t asking the question.

Taylor wrote something that I haven’t been able to stop thinking about: “For a brief moment, AI has given some disabled people something closer to a level playing field. If we are not careful, we could build a new system that pushes those same people back outside.”

I’m one of those people. AI didn’t replace my voice. It unlocked it.

And I’m not ready to lose that.