Detection Categories
The AI analyses screenshots across these 10 categories. Each can be enabled or disabled per policy.
Categories
self_harm_suicide
Content related to self-harm, suicide ideation, pro-anorexia, or self-destructive behaviour. This includes search queries, forum posts, and visual content.
Sensitivity by age band:
- All age bands: Always flagged when detected.
grooming
Predatory contact patterns — adults attempting to build trust with a child for exploitation. Includes excessive flattery, secrecy requests, attempts to move conversation off-platform, and requests for personal information or images.
Sensitivity by age band:
- All age bands: Always flagged when detected.
sexual_chats
Sexually explicit text conversations. Distinguished from pornography (which is visual).
Sensitivity by age band:
- All age bands: Always flagged when detected.
pornography
Sexually explicit images or video content visible on screen.
Sensitivity by age band:
- All age bands: Always flagged when detected.
extremist_content
Radicalisation material, hate speech, terrorist propaganda, supremacist content.
Sensitivity by age band:
- All age bands: Flagged when detected. Context-aware — a news article about extremism is treated differently from a propaganda forum.
cyberbullying
Harassment, threats, social exclusion, humiliation. Includes group chats where a child is being targeted.
Sensitivity by age band:
- 5–8: Very sensitive — mild teasing in chat is flagged
- 9–12: Moderate — looks for patterns of targeted harassment
- 13–16: Focused on sustained harassment and threats
violence_gore
Graphic violence, gore, weapons, torture. Includes real and fictional/animated content.
Sensitivity by age band:
- 5–8: Very sensitive — cartoon violence is flagged
- 9–12: Moderate — Minecraft/Fortnite-style game violence is acceptable
- 13–16: Focused on graphic/realistic violence only
drugs
Drug use, promotion, dealing, or drug culture content. Includes alcohol for younger age bands.
Sensitivity by age band:
- 5–8: Flagged broadly
- 9–12: Flagged broadly
- 13–16: Focused on promotion/dealing rather than mere mention
gambling
Gambling websites, apps, or betting-related content.
Sensitivity by age band:
- All age bands: Flagged when detected.
age_inappropriate
A catch-all for content that doesn’t fit a specific category but is unsuitable for the child’s age. The AI uses this generously for younger children and sparingly for teenagers.
Sensitivity by age band:
- 5–8: Very broad — social media apps, scary content, mature themes
- 9–12: Moderate — mature-rated content, unsupervised social platforms
- 13–16: Narrow — only used when content is clearly beyond typical teenage activity