OpenAI Launches Accessible Prompt Library for Nonverbal Communication Tools

AI · 4 min read

The library focuses on reliability and predictability when models are embedded in AAC systems—use cases where latency, hallucination, and ambiguous phrasing can have outsized negative effects. Prompts are organized by communicative intent (e.g., emergency, social niceties, medical clarification) and come with suggested safety filters and short-form templates for low-bandwidth devices.

Importantly, OpenAI provided developer guidance for multimodal fallbacks: when voice input is poor or images are unclear, the prompts recommend conservative paraphrasing and explicit confirmation steps. The toolkit also includes localization notes and a checklist for testing with neurodiverse users and speech-impaired participants.

Advocates welcomed the effort but urged continuous co-development with AAC communities. OpenAI said it will maintain the library as community-contributed artifacts and integrate anonymized feedback loops to improve prompt reliability over time.