AI Use & Limitations Policy

1. Scope

This policy applies to AI outputs generated through AskVerdict web, API, SDK, CLI, and partner integrations. It governs how users should interpret, validate, and operationalize generated content.

2. Core Principles

• AI outputs are advisory, not authoritative.
• Final accountability always remains with the user or customer.
• Higher potential impact requires higher verification and oversight.
• Safety, legality, and fairness must be checked before implementation.

3. Intended Use

AskVerdict is intended for structured reasoning use cases such as:

• Comparing alternatives and identifying tradeoffs
• Drafting arguments, plans, and decision memos
• Exploring uncertainty and opposing viewpoints
• Assisting policy or product discussions before human approval

4. Risk-Tiered Use Expectations

Low Risk

Brainstorming and internal ideation can usually proceed with light review and ordinary quality checks.

Moderate Risk

External communications, customer-impacting recommendations, or operational changes should include source validation and reviewer sign-off.

High Risk

Legal, medical, financial, employment, education, insurance, or safety-critical contexts require qualified human review and independent evidence before any action.

5. Human Oversight Requirements

• Verify material facts with primary sources before relying on outputs.
• Review assumptions, edge cases, and omitted alternatives.
• Require explicit approval for decisions with user, legal, or financial impact.
• Retain a decision record when AI materially informs a final outcome.

6. Verification Standards

• Cross-check factual claims with trusted primary sources.
• Validate dates, numbers, legal citations, and policy references.
• Test recommendations against real constraints and failure scenarios.
• Confirm output applicability to your jurisdiction and business context.

7. Prohibited Reliance and Prohibited Use

• Do not present outputs as licensed professional advice.
• Do not use outputs as the sole basis for irreversible decisions.
• Do not use AskVerdict to bypass legal, compliance, or audit duties.
• Do not use outputs to justify unlawful discrimination or rights violations.
• Do not deploy autonomous workflows that execute high-risk actions without human approval gates.

8. Sensitive Data and Prompt Hygiene

Enter only data needed for the task. If sensitive information is not required, remove or redact it before submission.

• Do not submit passwords, private keys, or access tokens.
• Avoid uploading sensitive personal data unless you have a lawful basis and required approvals.
• Use anonymized or pseudonymized examples during testing when possible.

9. Bias and Fairness Controls

AI outputs may reflect statistical bias or incomplete representation. You are responsible for checking potential adverse impact before applying outputs in people-related contexts.

• Evaluate whether recommendations affect protected groups.
• Require additional review where decisions involve eligibility, hiring, compensation, or access.
• Record rationale for final decisions where fairness risks exist.

10. Automation and Integrations

If you connect AskVerdict to external systems, implement approval checkpoints proportional to risk.

• Use read-only workflows by default for initial integrations.
• Add explicit approval for write actions that affect users or money.
• Log user approvals and execution context for auditing.

11. Incident Escalation

If you identify harmful, unsafe, or materially incorrect output that could create legal or security risk, pause use in that workflow and report the issue to your internal owner and to AskVerdict support.

12. Model and Product Changes

We may update models, ranking logic, safety filters, and product behavior over time. Output quality, style, and format may change as part of normal service improvements.

13. Related Policies

This policy should be read with ourTerms of Service,Acceptable Use Policy, andPrivacy Policy.

14. Contact

Contactsupport@askverdict.aifor policy clarification.