Pick one case from the AI Incident Database and write it up like an accident investigator: what happened, who was harmed, why, and what rule or practice would have prevented it.
Learning from incidents is how aviation and medicine became safe. AI policy needs the same habit, and each well-documented case helps.
Output: A one-page incident report.
Browse the AI Incident Database and pick a case with several news reports, so you can cross-check facts.
List what happened, when, and who was involved. Mark any facts that sources disagree on.
Separate the technical cause, the human and organisational causes, and missing safeguards.
Did any law or standard apply? Would the EU AI Act or your country’s rules have changed anything?
Propose one realistic change (technical, organisational or legal) and explain why it would help.
All steps done
Share what you made, then try another project. The PDF version has room for notes.
Next: Sort 20 AI news stories with the MIT AI Risk Repository Next steps
The MIT AI Risk Repository organises over a thousand AI risks into a shared taxonomy. Use it to classify 20 recent news stories, then look at what the pattern tells you.
If an employee at an AI company sees something dangerous, what protects them if they speak up? Compare protections across two or three jurisdictions.
Expand a handful of existing case studies into 25–40 profiles of legitimate and concerning customers, including historical cases of people who ordered dangerous biological material.
Will society respond to early AI accidents the way it should? Collect historical warning shots from other technologies and analyse which ones actually changed policy.
Pick one decision-maker, one AI risk and one decision they actually face. Lay out the options and make a clear recommendation in two pages.
