Researchers propose community-based audits to uncover AI harms to queer communities
A new paper on arXiv argues that standard AI audits miss harms to queer communities because they rely on aggregate data and fail to capture lived experiences. The authors say technical audits often overlook issues like erasure, misclassification, and discrimination because they lack community input.
Key points
- Workshop involved 16 queer community members auditing four AI case studies at EurIPS 2025
- Participants traced harms to problem definition, data collection, and exploitative contracts
- Authors argue standard audits fail to detect identity-specific harms due to lack of community input
The study describes a participatory workshop at EurIPS 2025 where 16 queer community members audited four AI case studies using the 4Cs harm taxonomy. Participants identified harms tied to problem definition, data collection, and contractual structures that exploit vulnerable groups. The findings suggest community-informed auditing could reveal harms that traditional methods overlook.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Society & Work
All →- Microsoft Copilot speculates Pudgy Penguins may partner with a major brand · 1 src
- Proton CEO Andy Yen canceled AI re-interviews after calling security testing 'loser thinking' · 3 src
- Anthropic’s Claude writes Linux exploit for unpatched CVE-2026-98432 · 1 src
- Hackers impersonate AI bots to hunt for credentials in credential-hunting campaign · 1 src
- Microsoft patents AI to edit live speech and faces in real time · 4 src
Comments
via GitHub Discussions