The same week one company announced an AI ASL avatar and another announced a set of rings that translate sign language in real time, I sat down and asked myself a question I think the rest of our community should be asking too. Who, exactly, is this for?
Not who pays for it. Not who builds it. Not who wins the press cycle. Who is it actually for?
Because if you look at the rollouts, the press releases, the demo videos, the breathless coverage from outlets that have never employed a Deaf reporter, the answer becomes pretty clear. These tools are for hearing people who want a frictionless way to feel like they have done their part. They are for institutions who want a checkbox. They are for product teams who want a launch event. The Deaf community is the prop, not the audience.
And I want to be careful here, because I do not think every AI tool aimed at accessibility is bad. Some of them are genuinely useful. Real-time captioning has improved my own life. Better speech recognition has opened doors. There are honest engineers working on honest problems and shipping work that makes daily life a little easier. I respect that work.
What I do not respect is the slow, polite substitution that is happening underneath it. The substitution of a real Deaf voice with a synthetic one. The substitution of a Deaf interpreter with a CGI puppet. The substitution of a Deaf reporter with a feed of articles that have been auto-translated by a machine that has never met a Deaf person.
This is the part I want to name out loud.
An Avatar Cannot Have a Stake
When a Deaf interpreter signs the news, they are not just translating words. They are translating from inside a lived experience. They know which signs the community uses this year and which ones we have moved on from. They know which topics deserve a heavier face and which deserve a lighter one. They know what it means when a school board cuts an ASL program, because they have probably watched it happen in their own city. They have a stake.
An avatar does not have a stake. An avatar cannot get angry about a story. An avatar cannot pause and shake its head. An avatar cannot decide that the official press release is missing the part that actually matters to our people. An avatar cannot raise an eyebrow that says here we go again.
The companies shipping these tools will tell you that fidelity will get better. That tomorrow’s model will have better facial expression, better discourse handling, better register. I believe them. The fidelity will improve. That is not the question.
The question is who gets to decide what is signed at all.
Whose Stories Make It Through the Filter
Here is what nobody is saying out loud. The minute a hearing-led company controls the AI ASL pipeline, they control the editorial layer too. They decide which articles get translated. They decide which press releases get the avatar treatment. They decide which voices get amplified and which ones quietly disappear.
And the incentives are not neutral. A company built to sell a translation product has every reason to translate the most pleasant, most uncontroversial, most advertiser-friendly content. They are not going to point their avatar at a story about cochlear implant overreach. They are not going to point it at a Deaf school closure that their own funder is connected to. They are not going to point it at a critical piece about their own product.
So what looks like an accessibility win on the surface becomes, in practice, a curation layer that decides what version of the Deaf world you get to see in your own language. That is not access. That is editorial control with extra steps.
I want to be plain. I do not think the engineers building these systems wake up wanting to censor our community. I think they are mostly well-intentioned. But intent does not protect us from outcome, and the outcome of letting a hearing-led pipeline decide what gets signed is that the boring, the safe, and the sponsorable will always beat the urgent, the complicated, and the real.
The Rings, the Avatars, the Hype
You probably saw the story about the AI rings that translate sign language in real time. Seven rings, motion sensors, a model trained on enough data to map hand shape and movement to text. The press loved it. It made the rounds because it scratches a very specific itch for a hearing audience — the fantasy that the Deaf community is one gadget away from being fully legible to them, on their terms, without any of them having to learn anything.
I am not going to tell you the engineering is fake. I am going to tell you the framing is.
The framing is that sign language is a code to be cracked. That if we can just get the model accurate enough, the work is done. That the friction in Deaf-hearing communication lives entirely in the Deaf person, and the solution is to make us machine-readable.
Anyone who has actually lived inside a Deaf community knows this is upside down. The friction does not live in our hands. It lives in the hearing world’s refusal to meet us halfway. No ring is going to fix that. No avatar is going to fix that. What fixes it is a generation of hearing people who learn to sign, hire Deaf staff, watch Deaf-made content, and treat us as peers in their own ordinary lives.
You cannot ship that as a product.
What Community-Sourced Actually Means
This is the part where I want to talk about what we are building, because the contrast matters.
DeafMonitor is community-sourced. That phrase gets thrown around loosely, so let me be specific about what it means here. It means every story that surfaces on the platform comes from a network of trusted publishers and individual contributors who are reporting from inside Deaf life, not from outside looking in. It means our editorial choices are made by Deaf operators, not by an algorithm trained to optimize for engagement. It means when something happens in our community — a school closure, a hiring win, an event, a controversy, a celebration — the people deciding whether and how to cover it are people who will see the consequences of that coverage in their own lives next week.
That is not glamorous. It is slower than an AI feed. It is messier. It will never be as cheap to operate as an avatar reading a wire service. But it is the difference between a community knowing itself and a community being narrated to.
I think that difference matters more than any model improvement that gets announced this year.
What I Am Not Saying
I am not saying we should boycott AI tools. I am not saying interpreters should refuse to work with technology. I am not saying the engineers at Sorenson, or at the ring startup, or at any of the other companies racing into this space, are villains. They are not.
I am saying our community needs to be very, very awake to who owns the pipeline. We have been here before. We watched the captioning industry consolidate into a small number of vendors with patchy quality. We watched the VRS industry get shaped by reimbursement rules that nobody in our community wrote. We watched interpreter agencies grow into businesses that often did not employ a single Deaf person in a decision-making role.
Every time, the tool was sold to us as a win for accessibility. Every time, the structural power ended up somewhere outside our community. Every time, we got the surface and somebody else got the leverage.
If we let the AI ASL wave happen without claiming our place in it, that is the version of history we are going to write again.
What Claiming Our Place Looks Like
Claiming our place means a few things.
It means insisting that Deaf people sit at the actual decision tables of any company shipping AI ASL products, not as advisors who get a polite hour, but as employees with hiring authority and product veto. It means demanding that training data for ASL models be licensed from Deaf signers, with Deaf consent, with Deaf compensation, and with Deaf-controlled review of the resulting model. It means refusing to let companies use the word community in their marketing until they can name the Deaf staff members on their team.
It means, for the rest of us, choosing where we get our news. Choosing where we get our information. Choosing whose interpretation of our world we trust. Choosing to support Deaf-led media even when it is a little less polished than the corporate alternative, because the polish is not what we are paying for. The point of view is.
And it means being patient with the version of community-sourced that we are building. DeafMonitor is not done. None of this is done. The point is not that we have arrived. The point is that we exist as a counterweight to the version of the future where an avatar smiles at us from a screen and reads us a press release written by a company that does not know our names.
The Quiet Test
If you want a quick test for any AI ASL product, here it is. Ask yourself, the next time one of these tools is announced, three questions.
Who on the founding team is Deaf? Who in the editorial layer is Deaf? And if our community had a serious complaint about how this tool was being used, would there be a single Deaf person inside the company with the power to do anything about it?
If the honest answer to any of those is no, you have not been shown an accessibility product. You have been shown a marketing product wearing the costume of one.
That is a hard line to draw in 2026. The pressure to celebrate every AI accessibility launch is enormous. The press will not draw the line for us. The funders will not draw it. The hearing institutions that consume these tools will not draw it.
So we have to draw it ourselves. Inside our own community. In our own words. From our own voices.
That is the work. That has always been the work. The rings and the avatars and the press releases are just the latest reminder of why it matters.
Let’s keep doing it.
— Bryan Leeper, Founder, DeafMonitor