Spot the AI's mistake
Check an AI-written triage summary against the evidence before anyone acts on it.
AI assistants can write a triage summary in seconds, and they sound sure even when they are wrong. Your job is to check the draft against the evidence before anyone acts on it. Each drill has the evidence and an AI draft. Flag every sentence you think is wrong. Some drafts have no mistakes at all.
Exploit strings at two web servers
WAF: Log4Shell exploit strings from one address
Not tried yet
Two cities, four minutes apart
Identity: impossible travel for s.kapoor
Not tried yet
A fake voicemail email
Mail gateway: reported phishing email, two clicks
Not tried yet
A locked account on the staff Wi-Fi
Identity: account lockout for j.thomas
Not tried yet
Files being renamed overnight
EDR: mass file rename on FS-02
Not tried yet
A leaked key starts new servers
Cloud audit: deploy key used from a new address
Not tried yet
Found something confusing or unrealistic?