AI Output and Legal-State Notice

Effective date: July 2, 2026
Last updated: July 4, 2026

1. Why this notice exists

Stetful is built around accepted company legal state: the reviewed record of what the company can support from evidence. In this notice, we refer to it as accepted state. The product loop is:

• Question / Event
• Evidence Coverage Check
• Answer / Confidence / Gap List
• Targeted Evidence Requests
• Optional Evidence Intake
• Optional Proposed State Impact
• Company Review
• Accepted Representation

This notice explains how to understand Stetful’s AI-assisted outputs within that loop.

2. Informational analysis, not legal advice

Stetful provides informational analysis based on available company legal records, evidence, and legal-state representations. It is not a law firm, lawyer, or substitute for counsel, and use of Stetful does not create an attorney-client relationship.

You are responsible for deciding when to involve qualified counsel.

3. Evidence coverage matters

A Stetful answer is only as strong as the available legal-state record, the evidence checked, the quality of that evidence, and any review or acceptance status.

When Stetful says it has not located evidence, that means the evidence was not located from the state and evidence checked for the relevant question or workflow. It does not necessarily mean the underlying legal act never happened.

4. Platform observations, customer reconciliation, interpretations, and accepted state

Stetful may refer to observed legal instruments, observed terms, platform observations, or observed assertions: limited evidence-backed records with confidence, time, visibility policy, and dispute status. They may be supported by evidence-basis records and provenance summaries.

A platform observation is not the uploaded artifact, not a confidential excerpt, not a tenant-specific interpretation, not a proposed representation, and not accepted state. Observed terms are structured parameters, not verbatim clause text by default. Customer reconciliation is the tenant-facing review layer for acknowledging, disputing, rejecting, or promoting an observation. A tenant interpretation is organization-specific reasoning about what evidence and platform observations mean for that organization. It becomes accepted state only through that organization’s authorized review and acceptance flow.

5. Confidence is not legal certainty

Stetful may use confidence or status labels such as supported, qualified, unknown, unsupported, contradicted, superseded, high, medium, low, or none.

Those labels reflect evidence coverage, source quality, consistency, recency, and review posture. They do not guarantee legal correctness or eliminate the need for counsel where legal judgment is required.

6. Proposed state is not accepted state

Stetful may identify proposed legal-state impacts, proposed representations, findings, action cards, or suggested next steps. These outputs are not accepted state just because Stetful generated them.

A proposed representation becomes accepted state only after an authorized reviewer accepts it through the product’s review flow.

7. Human review is required for reliance

Use human review before:

• relying on an output for a financing, diligence, customer, audit, board, regulatory, employment, privacy, security, or governance decision;
• sharing an output externally;
• making or accepting a legal representation;
• concluding that a record is missing, stale, contradicted, or sufficient;
• treating an output as legal advice; or
• deciding that counsel is unnecessary.

8. Privilege and confidentiality

Stetful treats customer legal records as confidential service data under the Terms and Privacy Policy. But using Stetful does not by itself create attorney-client privilege, attorney work product protection, or a lawyer-client relationship.

Privilege is fact-specific. Consult counsel for privilege-sensitive workflows.

9. AI providers and training

Stetful may use AI providers to process bounded information needed to provide the Service. Stetful does not use Customer Legal Content to train foundation models or third-party models.

Stetful uses AI providers only to provide the Service, not to train foundation models or third-party models. Provider human review of Customer Legal Content is not permitted except where legally or security-required or expressly enabled for a feature.

10. Good ways to use Stetful

Stetful is best used to:

• ask a specific company records question;
• see what the current company record supports;
• understand confidence, caveats, and uncertainty;
• identify the targeted record that would improve confidence;
• preserve review history; and
• make future answers stronger through reviewed, evidence-backed legal-state representations.

11. Bad ways to use Stetful

Do not use Stetful to:

• replace counsel where legal advice is required;
• make high-impact decisions without human review;
• treat AI-generated output as accepted state;
• claim a record is missing before coverage has been checked;
• upload records you are not authorized to provide;
• store unrelated sensitive data; or
• create external legal, investor, customer, audit, or board representations without appropriate review.

12. Contact

Questions about this notice: legal@stetful.com.