AI Incident Database - Incident 146
Source evidence cited by public records. This page shows where the source is used, its trust tier, and when it was last checked in the seed.
Records using this source
- minority-groups
entity | 4 June 2026 | 70%
minority-groups - organisation named in the AI Incident Database harm index.
- allen-institute-for-ai
entity | 12 Dec 2021 | 70%
allen-institute-for-ai - organisation named in the AI Incident Database harm index.
- Research Prototype AI, Delphi, Reportedly Gave Racially Biased Answers on Ethics
event | 12 Dec 2021 | 82%
A publicly accessible research model that was trained via Reddit threads showed racially biased advice on moral dilemmas, allegedly demonstrating limitations of language-based models trained on moral judgments.