AI Incident Database Updates

[Sean McGregor]

Hello All,

The Responsible AI Collaborative published our roundup of incidents for the months of December [0] and January [1] and added the ability to subscribe to incidents as they are added to the database [2].

We additionally presented a major update to our operational definitions of "AI Incident" at the NeurIPS workshop on Human-Centered AI [3] and our plans for investigating open source AI incidents [4] at the SafeAI conference earlier this month.

We have several more developments due out to the public shortly on the topics of language models, audit and risk assessment tooling, and expanding the number of people and organizations editing the database. Stay tuned!

The AI Incident Database is a collaborative project representing the hard work of many individuals and organizations. We invite your contributions in remembering the past so we can engineer a better future. Keeping up with the accelerating pace of AI incidents [5] requires collaboration systems at the scale of AI.

Best,

Sean McGregor

[0] https://incidentdatabase.ai/blog/incident-report-2022-december/

[1] https://incidentdatabase.ai/blog/incident-report-2022-january/

[2] https://incidentdatabase.ai/blog/user-story-spotlight-subscriptions-and-enewsletter/

[3] https://arxiv.org/abs/2211.10384

[4] https://arxiv.org/abs/2211.07280

[5] https://incidentdatabase.ai/summaries/incidentsOverTime