AI Transparency & Safety Notice

Sightra uses automated computer vision (YOLO object detection) and a large language model (Google Gemini) to produce spoken and written guidance. Outputs are generated automatically and may be wrong, incomplete, or delayed.

• People, vehicles, obstacles, signs, or surfaces may be missed or mislabelled.
• Distance and position descriptions are approximate, not measurements.
• Language may sound confident even when the system is uncertain.
• Network or server delay may mean the environment changed before you hear guidance.

• Do not cross roads or enter hazardous areas based on Sightra alone.
• Continue using your mobility training, cane, dog, and hearing.
• Stop and seek human help when unsure.

Volunteer or agent video calls are separate from automated AI. They may involve live audio/video through LiveKit. Only enable when you understand who may join and how sessions are moderated (operational policies apply).

Gemini processes structured detection data and optional text prompts according to Google’s terms and your API configuration. We do not use your data for public model training unless you separately opt in to any future secondary-use programme (disabled by default).