In Hong Kong, gender-based violence is considered a family affair and often goes unreported due to a lack of evidence. Studies show that the vast majority of incidents have one thing in common: a few very specific verbal insults tend to be shouted by perpetrators in the lead-up to physical abuse.

To combat this, we created Undercover, the first app designed to recognize these insults in any situation or context. When the AI hears these phrases at over 90 decibels, the app secretly starts recording the entire incident that follows, helping victims collect evidence of the abuse that is legally admissible in court. If the AI determines the incident is escalating, an emergency alert is sent to a pre-determined trusted contact.

The Undercover app runs in the background, designed to look generic to avoid detection, and is biometrically protected so there is no way for the abuser to access the recordings.