AI transparency sounds reasonable. But what happens when AI detection and watermarking systems cannot tell the difference between someone generating work with AI and a disabled person using AI as assistive technology?

In this video, I explain why AI watermarking and the EU AI Act have me concerned as a blind accessibility professional, app developer, person with ADHD, and heavy AI user.
AI helps me organize my thoughts, work with inaccessible information, understand images, read documents with OCR, create content, and communicate more effectively. For many disabled people, AI can function as assistive technology.

So what happens when a machine-readable signal simply says that AI was involved?
What happens when a blind person uses AI-powered OCR to read an inaccessible document?
What happens when someone with ADHD or dyslexia uses AI to organize their own thoughts?
What happens if an employer, school, publisher, or automated system sees an AI signal and assumes that person did not actually do the work?

I support transparency and protections against malicious uses of AI. But “AI was involved” does not automatically mean “AI did the work.”

Disabled people need to be part of this conversation while these policies and technologies are being designed — not after they create another barrier.

What do you think? Could AI watermarking unintentionally discriminate against disabled people?
If you think policymakers, AI companies, developers, educators, employers, and accessibility professionals need to hear this perspective, please share this video.