Can you spot the AI risk in 10 seconds?

Found this fun one: 10 real AI incidents, each hiding a specific governance risk. You get a few seconds to guess before the answer. Most have an actual name in AI governance.

A couple to try yourself:

- A company gives its AI assistant full autonomy to send emails, book meetings, and make purchases, no human approval needed. What risk is this?

- A hiring model performs great, but nobody documented where the training data came from. What risk is this?

- A customer sends a support email with hidden instructions buried in the text, and the company's AI assistant quietly follows them. What risk is this?

(Answers: excessive agency, data provenance, prompt injection, but there are 7 more in the video; watch the video to find out)

Full game here: https://youtu.be/bKMC_83Zr9A?si=zQGQd_QH00kCjjue?utm_source=reddit&utm_medium=organic&utm_campaign=incident_series&utm_content=62-ai-game

Question: how many out of 10 do you think you'd get?