AI Use Policy

Last updated: April 27, 2026 · Version 1.0

1. Scope

This policy describes how Statura Intelligence uses artificial intelligence to generate briefs, summaries, and analysis from public legislative data. It applies to every AI-generated artifact you see in the product: bill briefs, agenda summaries, lobbyist ROI explanations, brief drafts, and any future AI-assisted feature.

This policy is part of, and must be read alongside, our Terms of Service and Privacy Policy.

2. Models we use

Statura Intelligence uses third-party large language models. We do not train our own foundation models. Our current model providers and roles:

We may add additional providers (OpenAI, Google, open-source models) over time. We will update this table within 30 days of any production deployment of a new provider, and notify customers via email for material changes.

3. What we send to the AI

For each brief generation, Statura Intelligence assembles a context window from public-record sources only:

• The bill text as published by the Florida Senate or House (PDF or HTML)
• Sponsor metadata (name, district, party, filing history)
• Committee actions and floor votes from the official record
• Public sponsor statements, press releases, and official testimony
• Cross-references to companion bills, related sections of statute, and prior-session related bills
• Lobbyist comp filings + GAA appropriations lines (state-public records)

What we do not send to the AI:

• Your private workspace notes, custom keywords, or comments
• Your tracked-items list or priority flags
• Member emails, names, or any personal data of you or your team
• Your billing information
• Any data scoped to another customer's workspace

The only "your-data" inputs that ever reach the model are: (a) the bill or item you explicitly asked us to brief, and (b) optionally, your workspace's industry tags, used solely to weight relevance scoring within the brief output.

4. Training boundary

Your private content is never used to train AI models. Period.

Anthropic operates on a zero-retention policy for API customer queries: they do not log queries, do not train models on them, and do not retain content beyond the duration of the inference request. Our contract with Anthropic explicitly prohibits use of Statura Intelligence queries for any training, evaluation, or model-improvement purpose.

We periodically audit Anthropic's enterprise compliance posture as part of our sub-processor review. Their current policies are published at anthropic.com/legal/commercial-terms.

5. Citation guarantee

Every claim in an AI-generated brief is paired with a citation back to a public, primary source. We enforce this in two ways:

1. Generation-time: the model is prompted to cite every factual claim with a footnote tag. Briefs that fail the citation check are flagged and re-run.
2. Verification-time: a separate citation-resolution pass confirms each footnote URL is reachable, the cited content matches the claim, and the source is from our trusted publisher list (flsenate.gov, myfloridahouse.gov, *.fl.us, agency rule databases, etc.).

Briefs display an "AI Verified ✓" pill only when both passes succeed. The provenance drawer on every bill page shows the full audit trail: model used, generation timestamp, citations resolved, and any human edits.

6. Hallucination policy

Despite our citation guarantee, large language models can occasionally produce plausible-but-incorrect output. We treat this as a known failure mode and design around it:

• No uncited claims. If the model produces a sentence without a citation, the citation-resolution pass replaces or removes it.
• Bounded scope. Briefs are confined to facts derivable from the citation set. The model is instructed not to extrapolate beyond the source material.
• Confidence display. Every brief shows model + generation date prominently, so you can re-run if the underlying source has changed.
• Verification: we periodically sample briefs against the original sources and publish accuracy metrics (target: ≥99.5% citation correctness on a rolling 30-day sample).

You should always verify AI-generated content against the cited source before relying on it for legal, regulatory, or business decisions. The "AI Verified" pill confirms our citation-resolution pass succeeded. It does not certify the brief is free of error.

7. Human oversight

Statura Intelligence is an "AI-assisted" tool, not a fully autonomous agent. Humans remain in the loop at every consequential step:

• Brief generation is user-initiated. We don't generate briefs without an explicit request.
• Edits to a brief are tracked and shown in the audit trail. Any team member with permission can modify generated content; the original AI version is retained.
• Alerts are based on deterministic rules (stage change, keyword match), not AI judgment.
• Tracked-item priority is set by you, not the AI.
• Subscription and billing actions are never AI-initiated.

8. Opt-out

Workspace administrators can disable AI features for their workspace at any time in Workspace settings → Tracking preferences. Disabling AI:

• Stops new AI brief generations
• Removes the "Generate brief" button from bill pages
• Preserves existing briefs (does not retroactively delete)
• Does not affect deterministic features (alerts, tracking, ROI dashboards, agenda watch)

You can also delete individual AI-generated briefs at any time. Deletion is permanent after the 60-day workspace retention window.

9. AI output is not legal advice

Statura Intelligence briefs and analysis are informational only. They are not:

• Legal advice
• Lobbying registration or compliance advice
• A substitute for review by a licensed attorney or registered lobbyist
• An authoritative interpretation of Florida statutes or rules

The official source (the bill text, the committee record, the agency rule) is always authoritative. We link to it from every brief. Use AI output as a starting point, not an endpoint.

10. Disclosure to your clients

If you are a registered Florida lobbyist or government-affairs professional using Statura Intelligence output in client-facing work product (e.g., quarterly client memos), we recommend disclosing AI assistance in line with your firm's professional-responsibility policies and any applicable bar guidance on AI-assisted work.

Statura Intelligence provides every brief with a clear "AI-generated" indicator and a citation list. You are responsible for whether and how you forward AI-generated content to clients.

11. Updates to this policy

We will update this policy whenever we materially change how AI is used in the product: new providers, new feature scope, changed retention policies, or revised opt-out controls. Material changes are notified by email and posted with a new "Last updated" date.

Questions, audit requests, or specific compliance asks: legal@statura.app.

— END OF AI USE POLICY —