AI + Sign Language: What We Learned from Recent Research (2023–2025)

Over the last three years, AI and sign-language research has grown very fast. Many people think AI can “translate sign language,” but the real picture is more complex.

We in the Deaf community have been discussing these issues, and many of us share the same concerns about how AI handles our sign languages.

Here are the key points from the latest research:

• AI is getting better at creating signing videos with facial expression, body movement, and timing.

• New datasets (including 72+ hours of Chinese Sign Language) help move from single signs to full sentences.

• But AI still makes many mistakes with lighting, skin tone, facial grammar, and dialect differences.

• Reviews show AI is not reliable for hospitals, police, legal systems, emergencies, or education.

• Deaf-led organisations warn that AI must not replace interpreters.

• BSL pragmatics, intonation, expression and regional dialects are still too complex for AI to understand.

• Researchers say many datasets lack diversity and may erase dialects or culture.

  
  

 Main message:

AI can support sign languages, but it cannot replace them.

The future must be Deaf-led, ethical, safe, and respectful of all sign languages.

I may make a BSL video version soon, because BSL and English are not the same. Sign-language communities deserve this information in their own languages.