TAI Hackathon — Problem Statement 5

We developed a single Jupyter notebook that turns the AI Incident Database into interactive visualizations and a shareable web app. No coding required — just press play.

Incidents from 1983–2026 · 3 taxonomies · 12 charts · 1 web app

1. Open TAI_AIID_Research_Notebook.ipynb in Google Colab
2. Set runtime to T4 GPU (Runtime → Change runtime type → T4 GPU)
3. Click Runtime → Run all
4. Wait ~2 minutes. Done!

• 12 interactive visualizations — incident growth, harm categories, company rankings, geographic map, failure patterns, emerging risks, and more.
• 6-tab Gradio web app — searchable, filterable, shareable via a public link and an AI-Assistant.
• Unified dataset — three expert taxonomies (MIT, GMF, CSET) merged into one master table.

• Data: AI Incident Database — CC BY-SA 4.0
• Notebook code: MIT License

Team Undefined — TAI Hackathon 2026, Problem Statement 5

Members:

1. Jugal Gajjar
2. Kamalasankari Subramaniakuppusamy
3. Kaustik Ranaware

Making AI Harms Visible, Researchable, and Actionable.