AI failures are not just technology failures.
Behind AI incidents are decisions made by people, organisations and sometimes systems acting without meaningful human oversight.
Headlights is an independent library of 234 documented AI incidents from around the world. I examine what happened, how people and organisations responded, and what the public record tells us about the behaviours involved.
When you look beyond the technology, some interesting patterns begin to emerge.
Start with the evidence. Then ask what people did.
The library is here to be used, questioned and built upon.
DOI: 10.5281/zenodo.23312145
Headlights
I studied criminology and volunteered with victims of crime. One thing stayed with me: if you want to understand what went wrong, you need a reliable record of what happened.
I then spent more than twenty years working in enterprise technology, including with organisations responsible for critical infrastructure.
I became interested in a question that technology discussions often overlook.
What happens around the AI system?
Who questions it? Who trusts it? Who notices when something is wrong? And what happens when nobody does?
So I started researching real incidents, one at a time.
Headlights is my attempt to understand those questions through the evidence, and make that research available to anyone who wants to examine it.