Trusted enough for official use?
What AI translation needs before it can carry a government's name
The deployment of large language models (LLMs) for multilingual public service delivery is frequently promoted as a cost-effective solution for language access. However, official use demands a much higher threshold of reliability.
For an AI-assisted language service to be trusted by public professionals, it must incorporate formal institutional assurances. Machine-translated output cannot carry a government department's authority without explicit human-in-the-loop review, quality disclosures, and robust data confidentiality safeguards. Our research investigates how these conversion factors causally shape professional trust.