Advisory by Architecture: Why Dawn Will Never Predict People

Every design decision in Dawn Intelligence exists to prevent one thing: the prediction of individuals.

This is not a policy. It is not a charter that can be revised at the next board meeting. It is enforced at the code level, in the request-gating layer that sits between every API call and every output the platform produces.

Dawn fuses seven signal sources: police recorded crime, news media, charity and third-sector reporting, Google Trends, forecasting models, macro events, and derived signals. It covers every police force in England, Wales and Northern Ireland. It processes ten years of historical data across 90+ Home Office offence codes and 8 safeguarding categories.

None of this is pointed at people.

The platform operates at the level of the institution, the region, the pattern. It tells a safeguarding lead that domestic abuse referrals in their area are rising faster than the national trend. It tells a police commissioner that retail crime in their force area correlates with night-time economy density in ways that other forces have already acted on. It tells a local authority that three of their neighbouring boroughs have seen the same signal shift, six months earlier.

It does not tell anyone who is at risk. It does not score individuals. It cannot be turned into a surveillance tool, because the architecture does not permit it.

This is the only position that survives the next decade of UK AI regulation. The Data (Use and Access) Act 2025, the AI Security Institute framework, and the direction of travel from every serious regulator in Europe all point the same way: away from individual prediction, towards institutional intelligence.

Dawn is already there. Not because we saw this coming. Because it was the only ethical way to build it in the first place.