1
Enable content safety filters
Foundry includes Azure AI Content Safety filters by default. Check the project safety policy — Severity Threshold should be Medium for Hate, Sexual, Violence, Self-Harm.
All labs · Lab 5 of 5 · 1-2 hours
Responsible AI, Cost & Audit
Apply content safety, monitor token usage, and review the audit trail.
Prerequisites
- At least one of the previous labs completed
- Access to the team's Foundry project metrics page
Steps
2
Check token usage in the project metrics
Open Monitoring > Token usage. Filter by your deployment. Note prompt_tokens vs completion_tokens for the last 24 hours.
3
Set a daily budget alert
From the Cost Management blade, set a 1,800,000 KRW monthly alert. This is the operations-team default — verify the email recipient.
4
Review the audit log
Use the audit log page to confirm role assignments and the last login of every team member. Tutors must see four logins per team.
5
Document the risk register
Write 5 lines in your README: prompt-injection risk, hallucination risk, privacy risk, cost risk, and the mitigation for each.
Verification
When all three (or more) of these are true, you can mark the lab complete and move on.
- Content safety filters are at Medium or stricter
- Token usage chart shows the team's last 24 hours
- Budget alert is configured to the operations email
- Risk register section exists in the team README
Azure cost
These checks are monitoring-only and have no per-call cost.
Responsible AI
The risk register is the most important deliverable here — judges and tutors will read it.
Hackathon connection
Sample repo — risk register template