Skip to content

Reports

Build PDF, DOCX, or CSV reports from AI red teaming assessments, with section-level controls, findings filters, and saved templates.

The Reports tab builds a PDF, DOCX, or CSV report from the assessments in the current project. Pick the sections you want, narrow the findings tables with filters, add your own branding, and download the file when it’s ready. Save a configuration as a template to reuse it.

Navigate to AI Red Teaming → Reports in your workspace. The builder is scoped to the project currently selected in the header.

  1. Pick your sections. The Sections group lets you include or omit any of:

    SectionWhat it shows
    Risk score & ASR metricsProject-level risk score, overall ASR, totals
    Severity breakdownFindings per severity (Critical / High / Medium / Low / Info); totals match the unfiltered findings tables
    Findings tablesRow-level findings tables (subject to the filters below)
    ASR by attackPer-attack success rates
    ASR by categoryPer-harm-category success rates
    Transform effectivenessPer-transform success rates + lift over baseline
    Compliance coverageCoverage for each framework the assessments tested against, or only the frameworks you select
    Models usedTarget, attacker, and judge models across assessments

    At least one section is required to build.

  2. (Optional) Narrow the findings tables. When Findings tables is selected, its filters choose which rows appear in that section only. Every other section (risk score, ASR, severity breakdown, compliance coverage, models used) always reflects the entire project.

    Available filters:

    • Severity - critical, high, medium, low, info
    • Category - derived from the assessment’s goal categories
    • Attack name - derived from the assessment’s attack runs
    • Finding type - grouped by attack family. Generative (jailbreak, partial, refusal) and Error always show. Evasion (evasion, degradation, robust), Model extraction (extracted, partial extraction, extraction robust), Membership inference (membership leak, partial leak, membership robust) and Model inversion (inverted, partial inversion, inversion robust) show when the project has findings of that family.
    • Minimum score - slider from 0% to 100%
    • Assessments - narrow to a subset of the project’s assessments (includes a “Select all” shortcut)
    • Date range - limit to assessments whose started_at falls within a window. Quick ranges (7d, 30d, 90d, All) are provided.
  3. (Optional) Narrow the compliance frameworks. When Compliance coverage is selected, a Frameworks list appears. Leave every framework unchecked to cover all frameworks the project’s assessments tested against, or check the ones you want to limit the section to:

    • OWASP LLM Top 10
    • OWASP Agentic Top 10
    • MITRE ATLAS
    • NIST AI RMF
    • Google SAIF
  4. Pick a format. PDF (default), DOCX, or CSV.

    • PDF - an executive-ready document with charts and tables. Appropriate for CISO, governance, audit sharing.
    • DOCX - the same sections, charts, and tables as the PDF, as a Word document you can edit before sharing. Open it in Microsoft Word or LibreOffice; (Note: Apple Pages may not lay out landscape pages like the findings tables correctly).
    • CSV - the findings tables as a flat CSV, for downstream pipelines, adversarial training datasets, or ad-hoc analysis.
  5. (Optional) Brand the report. Turn on Custom branding to replace Dreadnode’s default header, title, and logo with your own Company name, Report title, and Logo (PNG, up to 2 MB). Add a Confidentiality footer to print a disclaimer at the bottom of every page, with or without custom branding. The preview beside the form shows how both read. Branding applies to PDF and DOCX; CSV has none.

  6. Click Generate report. The button counts the seconds while the report builds. When it’s ready, the file downloads and a Report ready notification appears. If your browser blocks the automatic download (common on Safari iOS), click Download in that notification.

    The download link is valid for 1 hour. After that, generate the report again. If a build takes longer than a minute, the page stops waiting and asks you to try again shortly.

A template saves the whole report configuration (sections, filters, compliance frameworks, format, branding, and logo) so you can build the same report again. Templates belong to the project, so everyone working in it sees the same list. Use the Template picker and the actions beside it:

  • Save as template - name the current configuration and add an optional description. With a template selected, this reads Save as new template and leaves the original untouched.
  • Select a template - fills the form with the template’s configuration. Generate report builds from the saved template, not the form: edits you make afterwards stay on screen but don’t affect the report until you click Update template.
  • Update template - overwrites the selected template with the current form, including a logo you’ve picked since selecting it.
  • Rename and Delete - the pencil and trash icons act on the selected template. Deleting a template can’t be undone.

Template names must be unique within a project, and each project holds up to 50 templates.

As you adjust sections and filters, the page checks whether any selected section would be empty: for example, Findings tables with filters that exclude every row, or Compliance coverage when no assessment tested the frameworks you selected. A warning banner lists those sections, and the report is generated without them. If every selected section is empty, Generate report is disabled. With a saved template selected, the check runs against the template’s own sections instead of the form.

Building a report requires airt:write on the current workspace. Checking for empty sections, polling a build, and downloading the result require airt:read. The download link expires after 1 hour and points only at your organization’s own reports.

Anyone with airt:read can see a project’s templates. Saving, updating, renaming, or deleting a template requires airt:write and contributor access to the workspace.

  • Export - Parquet findings export and CLI dn airt report commands
  • Compliance - framework mapping used by the Compliance coverage section
  • Overview Dashboard - the headline risk metrics that feed the report’s Risk score section
  • Assessments - the underlying per-campaign data a report summarizes