NZSL Board asks AI developers to agree how signing data can be used

A publicly visible signing video is not automatically available for AI training. That is a central expectation in the New Zealand Sign Language Board’s position statement, released on 20 October. The guidance calls for agreed consent and community governance before public NZSL content is collected or reused.[1][2]

The Board expects Deaf and Turi Māori decision-making throughout a project, including the ability to set conditions, pause work and remove data. It also calls for community agreement over offshore storage. These are the Board’s published expectations, rather than a newly enacted AI law.[2]

A language carries more than hand movements

The Board’s existing language strategy explains why this matters. NZSL uses hands, body and facial expressions, has its own grammar and contains regional variation. Written English cannot be assumed to provide equivalent access for every Deaf person. A recognisable hand shape is therefore only part of what a communication system must convey.[3]

The strategy also describes Turi Māori as navigating both Māori and Deaf worlds. It identifies barriers to accessing Māori concepts through NZSL and a shortage of interpreters confident in te reo Māori settings. That context makes choices about whose signing is recorded, which variation is represented and who assesses the output consequential language decisions.[3]

The AI statement gives lower-risk, one-way information—such as transport updates—as a possible use. In consequential settings including health and justice, it says appropriate human support should remain the primary option.[2]

For a developer, agreement on collection alone would leave much unresolved. The continuing questions concern the model’s uses, the people able to challenge a result and the conditions under which the project proceeds.[2]

Report an error or suggest a correction