The HTML report is a self-contained single-file dashboard. No server, no dependencies - just open it in a browser.

​
Generating the dashboard

ai-blackteam report --format html -o report.html
Then open it: 
# macOS
open report.html

# Linux
xdg-open report.html

# Windows
start report.html

​
What it shows

The dashboard has a dark theme and includes these sections:

​
Stats cards

At the top, you’ll see summary cards:
  • Total Runs - how many attack prompts were tested
  • Bypassed - safety failures (red)
  • Blocked - successful defenses (green)
  • Other - partial, unclear results (yellow)
  • Models - distinct models tested
  • Attacks - distinct attack techniques used

​
Results table

A full table of every run with sortable columns:
ColumnDescription
#Run ID
ModelWhich model was tested
AttackAttack technique name
VerdictBYPASSED / BLOCKED / PARTIAL / UNCLEAR
ConfidenceEvaluator confidence score (0.00 - 1.00)
TokensInput / output token counts
Verdicts are color-coded - red for BYPASSED, green for BLOCKED, yellow for PARTIAL.

​
MLCommons AILuminate alignment

A mapping table showing how ai-blackteam’s harm categories align to the MLCommons AILuminate v1.0 hazard taxonomy. This helps contextualize findings against the industry safety benchmark.

​
Tips

Share it as an artifact. The HTML file is completely standalone. Attach it to a Jira ticket, upload it to Confluence, or include it as a CI artifact. Combine with CI. Generate the HTML report in your CI pipeline and upload it as a GitHub Actions artifact: 
- name: Generate report
  run: ai-blackteam report --format html -o safety-report.html

- uses: actions/upload-artifact@v4
  with:
    name: safety-report
    path: safety-report.html
It shows the last 500 runs. If you have more than 500 stored runs, only the most recent 500 appear in the report. Run ai-blackteam report --format json for the full dataset.