AI & Safety Standards
How WhatNow? uses AI to assist human decision-making without allowing artificial intelligence to hallucinate emergency procedures.
What AI Does
- •Understands Natural Language: Interprets everyday descriptions like "someone took my backpack on the train".
- •Identifies Situations: Matches user descriptions to 52 structured candidate scenarios.
- •Personalizes Context: Tailors explanations based on location (e.g. travelling overseas).
- •Conversational Assistance: Reassures users and breaks complex recovery timelines into small, manageable steps.
What AI Does NOT Do
- •No Invented Procedures: AI is prohibited from inventing emergency steps or legal requirements.
- •No Fake Helpline Numbers: Emergency numbers and official links come strictly from verified registries.
- •No Invented Deadlines: Statutory timeframes come from verified sources; variable deadlines are explicitly labeled.
- •No Medical or Legal Advice: WhatNow? does not replace emergency dispatch, lawyers, or medical doctors.
Why This Architectural Boundary Matters
In high-stress situations, confident-sounding incorrect information can be dangerous. Standard large language models (LLMs) can hallucinate plausible-sounding government URLs or outdated phone numbers.
WhatNow? deliberately separates AI Intent Classification from Authoritative Emergency Data. The AI interprets what you need, but the actual instructions, links, and helplines are rendered deterministically from curated, audited WhatNow datasets.
User Responsibility & High-Stakes Situations
If you or someone else is in immediate physical danger, medical crisis, or active crime scene, do not rely solely on web platforms. Contact your local emergency services (112 / 911 / 999 / 1930) immediately.