Artificial Intelligence Incident Database

Website developed by the Partnership on AI. The AI Incident Database is a collection of harms or near harms realized in the real world by the deployment of intelligent systems.

From the Website:

You are invited to submit reports to the database, whereupon accepted incidents will be indexed and made discoverable to people developing and deploying the next generation of AI technology to the world. Artificial intelligence will only be a benefit to people and society if we collectively record and learn from its failings.

Why “AI Incidents”?

Intelligent systems are currently prone to unforeseen and often dangerous failures when they are deployed to the real world. Much like the transportation sector before it (e.g., FAA and FARS) and more recently computer systems, intelligent systems require a repository of problems experienced in the real world so that future researchers and developers may mitigate or avoid repeated bad outcomes.

What is an Incident?

The initial set of more than 1,000 incident reports have been intentionally broad in nature. Current examples include,

  • An autonomous car kills a pedestrian
  • trading algorithm causes a market “flash crash” where billions of dollars transfer between parties
  • facial recognition system causes an innocent person to be arrested

You are invited to explore the incidents collected to date, view the complete listing, and submit additional incident reports. Researchers are invited to review our working definition of AI incidents.

Current and Future Users

The database is a constantly evolving data product and collection of applications.

  • Current Users include system architects, industrial product developers, public relations managers, researchers, and public policy researchers. These users are invited to use the Discover application to proactively discover how recently deployed intelligent systems have produced unexpected outcomes in the real world. In so doing, they may avoid making similar mistakes in their development.
  • Future Uses will evolve through the code contributions of the open source community, including additional database summaries and taxonomies.

When Should You Report an Incident?

When in doubt of whether an event qualifies as an incident, please submit it! This project is intended to converge on a shared definition of “AI Incident” through exploration of the candidate incidents submitted by the broader community.

Learn more about the database, or read about it on the PAI BlogVice NewsVenture Beat, and arXiv among other outlets.