AI Policy

Purpose and oversight

AI may assist with evidence collection, classification, analysis, scenario evaluation, drafting, and workflow coordination. Human and governed oversight should reflect the significance, reversibility, and risk of the decision or action.

Aurora does not permit an AI system to make an autonomous binding commitment, spend funds, change protected production systems, or conduct external outreach without the authority and controls required for that action.

Models and providers

Aurora may use OpenAI, Anthropic, OpenRouter, and other approved model or infrastructure providers where appropriate to an authorized service. Provider selection can vary by capability, availability, risk, contract, and customer requirements.

Use of a provider does not imply endorsement of every provider service or permission. Provider terms, data practices, and technical limits apply to their independent processing.

Data minimization and authorized processing

• Use only information relevant to the authorized purpose.
• Avoid entering credentials, secrets, or unnecessary sensitive information into model prompts.
• Respect customer permissions, confidentiality, privacy, intellectual property, and provider policies.
• Apply additional restrictions when information or outcomes carry elevated legal, financial, employment, safety, or reputational impact.

Evidence, provenance, and limitations

Aurora seeks to preserve evidence, provenance, assumptions, and decision traceability where relevant. AI output can still be incomplete, biased, probabilistic, outdated, or incorrect.

Outputs must be evaluated against reliable evidence and the operating context. A generated answer is not a substitute for required professional advice or accountable human judgment.

Prohibited uses

• Unlawful discrimination, manipulation, deception, impersonation, or surveillance.
• Unauthorized scraping, credential use, access, or evasion of platform controls.
• Generation or execution of malicious code, fraud, abuse, or rights violations.
• Misrepresenting AI output as verified evidence or concealing material limitations.
• Using AI to bypass required human, legal, security, or corporate approval.

Evaluation and contact

Aurora evaluates AI behavior and provider suitability as capabilities and risks evolve. Evaluation does not guarantee that every output will be correct or that every risk can be eliminated.

Questions about this policy: contact@auroraos.online.