# Publish Trust Reports That Reviewers Accept

Turn evaluations and live monitoring into audit-ready proof for security reviews, RFPs, and renewals.

Trusted by teams building AI help desks, copilots, and agents.

[Contact us](/contact)

## Why It Matters

- **Reviews Ask The Same Questions Every Time.** Stakeholders want scope, methods, thresholds, results, and ownership in one place. Swept condenses evaluations and supervision data into a report reviewers can scan quickly, with links to the underlying evidence.
- **Production Oversight That Prevents Surprises.** Sample the right traffic, lock a baseline, track deltas, and route alerts with context. Every report reflects the latest status and the history behind it.

## How Swept AI Certification Works

1. **Select What To Summarize.** Choose an evaluation run or a live period, including relevant models, prompts, datasets, and environments.
2. **Add Controls & Ownership.** Attach security and process controls, list owners and escalation contacts, and link incidents and mitigations.
3. **Generate the Trust Report.** Create a readable report with scope, thresholds, methods, and outcomes, plus appendix sections for evidence.
4. **Share and Track.** Share a private link with access controls, request a review, track comments, and sign off. Export a PDF if required.

## Inside a Trust Report

- **Overview.** Purpose, scope, models, prompts, datasets, environments, and date ranges.
- **Methods.** Tasks, graders, metrics, thresholds, sample sizes, and baselines.
- **Controls.** Data handling, privacy, change management, incident response, and responsible AI notes.
- **Ownership.** System owners, reviewers, escalation contacts, and version history.
- **Results.** Accuracy, hallucination rate, safety flags, bias indicators, latency and cost, and pass or fail against thresholds.

## Sharing and Governance

- Roles and permissions for who can change thresholds and approve fixes.
- Watermark options, link expiry, domain allow lists, and SSO enforcement.
- Redaction rules for sensitive content in shared examples.

## Compliance Workflows

- **Framework Mapping.** Align tests and controls to common frameworks, for example SOC 2 topics and ISO-style categories, without implying certification.
- **Reviews and Approvals.** Assign reviewers and capture comments and decisions with timestamps.
- **Checklists and Gates.** Define required sections and evidence, and block report publishing until items are complete.
- **Renewals and Changes.** Clone a previous report, highlight deltas since last approval, and keep history for audits.

## Works with Your Stack

50+ integrations and counting, including OpenRouter, OpenAI, Anthropic, Gemini, Ollama, Mistral AI, Vercel AI SDK, Zendesk, Helpscout, and Fin.

## FAQ

- **Is this a compliance certification?** No. Swept produces internal trust reports and evidence packages that support your compliance efforts. We do not issue certifications. Reports can map to framework categories for easier reviewer handoff, but they represent your own evaluations and monitoring, not third-party audits.
- **Can we restrict who sees a report?** Yes. Reports have access controls including private links, domain allow lists, SSO enforcement, and link expiry. You decide who can view, comment, or approve, and can revoke access at any time.
- **How often should we refresh a report?** Most teams refresh after significant model or prompt changes, before major reviews, or on a regular cadence for renewals. Swept tracks version history so reviewers can see deltas since the last approval.
- **What if reviewers want raw evidence?** Reports link to underlying evaluation runs and supervision data. Reviewers with appropriate permissions can drill into specific examples, grader outputs, and metric breakdowns. You can also export CSV or JSON for offline analysis.