Tim Scannell, Accessibility Consultant:

Enhancing Access for the Deaf and Hard of Hearing by Focusing on Sign Language, Digital Accessibility, and Information Access for Everyone.

Under the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD) , States, public bodies, universities, and organisations have clear obligations to recognise, promote, and protect sign languages — and the people who use and sustain them. As artificial intelligence (AI) technologies increasingly enter the field of sign language translation and accessibility, they are often framed as progress: faster access, wider reach, lower cost. Accessibility is impo

Policy Summary This article sets out key considerations for responsible investment and deployment of AI for sign-language communication, particularly British Sign Language (BSL). Key points for policymakers, funders and commissioners: Current AI systems do not yet produce fully natural BSL and should not be treated as equivalent to human interpretation. Accuracy, safety and trust are critical in high-risk sectors such as healthcare, justice, education, employment and publ

Alt Text:: Illustration showing a busy public transport environment during an emergency. Deaf and hearing passengers, staff, and a driver communicate using multiple methods: sign language, written text, mobile alerts, smartwatches with haptic warnings, and face-to-face interaction. A sign language avatar appears on a screen, but the focus is on real people communicating directly. The image emphasises that accessibility is not tokenistic or limited to one avatar, but requires

“Real-time generative AI” is often misunderstood. Real-time = fast response Generative = creates output AI = automated decision-making software Put together, real-time generative AI means: ➡️ The system can generate content immediately when triggered. What it does not automatically mean: ❌ It understands language ❌ It interprets meaning or intent ❌ It handles live, free-form communication ❌ It performs true translation ❌ It manages unexpected situations Speed ≠ understan

When AI meets sign language, avatars are often the first thing we see. They’re visual, impressive, and easy to demonstrate. But if AI support for Deaf communities stops at avatars, we’ve misunderstood the real need. An avatar is a delivery layer , not intelligence. Infographic showing that sign language AI is more than avatars, highlighting the need for two-way communication, understanding context, and working everywhere to support real accessibility. True AI assistance for s

A false mental model has formed around “AI sign language translation.” Many people now believe that public systems (such as station screens) provide real-time sign language translation. But this belief does not match how most of these systems actually work. This false mental model has formed because: The word “real-time” is used The label “AI” is applied Signed output appears instantly and looks like language An infographic contrasting the false belief that AI provides live

Sign-language AI is often discussed as a single capability, but in practice, it encompasses very different types of tools , each with distinct levels of maturity, risk, and impact. Before funding or adopting any solution, it’s important to be clear about: What type of system is actually being proposed whether it supports one-way translation (e.g. text or speech → signed output) or claims bi-directional translation (signed input ↔ text or speech) What remains human-led ver

I reviewed current sign-language AI tools to understand what they can really do, what they cannot do, and where risks are being created for Deaf users and organisations. The main issue Many very different AI tools are being described as “sign-language translation” , even though they do very different things. This creates confusion, unrealistic expectations, and poor accessibility decisions. Three types of sign-language AI tools Across products, demos, and research projects,