AI Policy Terms

1. What We Mean by AI

In this Policy, "AI" means software that generates, summarizes, classifies, ranks, recommends, predicts, extracts, or helps act on information using machine learning, large language models, foundation models, automated rules, retrieval systems, tool-calling agents, or similar technology.

Snug uses AI to make benefits easier to understand and administer. AI may appear as an AI tool, in-product assistant, plan comparison explanation, recommendation, document workflow, text-selection explanation, search answer, or another automated feature.

2. How Snug Uses AI

Snug may use AI to:

• Explain benefits concepts, including ICHRAs, Section 125 plans, voluntary insurance, marketplace plans, plan attributes, coverage terms, and enrollment steps.
• Help employers compare benefit strategies, budgets, employee contribution approaches, voluntary products, census information, and setup options.
• Help employees compare health plans, understand ICHRA allowances, review voluntary insurance, check plan details, and navigate enrollment.
• Search, summarize, and explain Snug documentation or user interface context.
• Extract or organize information from user inputs, including medication names, plan preferences, benefit interests, or employer setup answers.
• Rank or compare plans based on plan data, user inputs, employer contributions, estimated costs, drug coverage, provider coverage, metal level, plan type, quality, and other benefit-related factors.
• Generate drafts, summaries, recommendations, UI cards, tables, next steps, or support handoff notes.
• Assist support, compliance, operations, product improvement, safety monitoring, abuse prevention, and quality review.

AI output may be personalized based on information you provide, information available through your employer or employee account, plan data, public or government data sources, third-party benefit data sources, page context, prior conversation context, and other information available through the Services.

3. What AI Does Not Do

Snug's AI tools do not replace professional advice or final user review.

Unless Snug expressly states otherwise in a separate signed agreement or required disclosure, AI output is not:

• Legal advice.
• Tax advice.
• Medical advice, diagnosis, or treatment advice.
• Investment, financial planning, accounting, payroll, fiduciary, or ERISA legal advice.
• Underwriting advice.
• Claims advice.
• A guarantee that any plan, benefit, reimbursement, election, tax treatment, coverage, provider, drug, premium, subsidy, affordability result, compliance status, or enrollment outcome is available, accurate, final, or approved.

You are responsible for reviewing information before relying on it. Employers remain responsible for benefit decisions, plan sponsorship obligations, employee notices, payroll practices, tax treatment, and compliance obligations. Employees remain responsible for reviewing plan information, confirming provider and drug coverage, and making their own enrollment selections.

4. Human Review and User Confirmation

Snug may allow AI to explain options, prepare recommendations, draft summaries, or help complete workflows. However, AI will not intentionally complete the following actions without user confirmation or another approved authorization step:

• Final employee plan enrollment or plan selection.
• Employer benefit activation.
• Signing or approving legal, benefit, plan, or tax-related documents.
• Applying employer recommendation changes to active benefits.
• Changing employee records in a material way.
• Sending enrollment invitations where confirmation is required by the product flow.
• Submitting applications, binding coverage, cancelling coverage, or changing coverage.
• Initiating payments, payroll deductions, reimbursements, or other financial transfers.

Snug may still use automated systems to validate, process, route, record, or execute an action after you confirm it.

You may request human review or support when available through the Services. Snug may also route an issue to a human reviewer when AI is uncertain, when information is missing, when a question is outside AI's permitted scope, or when a safety, privacy, compliance, or quality issue is detected.

5. Accuracy and Limitations

AI systems can be inaccurate, incomplete, outdated, or misunderstood. Plan data, provider directories, drug formularies, premiums, subsidies, employer contributions, benefit terms, and legal requirements can change. AI may also misunderstand your question, your context, or third-party data.

Snug works to reduce those risks by using safeguards that may include:

• Account and role-based access controls.
• Tenant scoping for employer and employee data.
• Data minimization in prompts and tool calls.
• Source-grounded answers where available.
• Structured output validation.
• Product allowlists and schema validation.
• Rate limits and abuse controls.
• Privacy-preserving AI usage logs.
• Prompt-injection defenses.
• Testing and monitoring for unsafe, inaccurate, or out-of-scope responses.
• Human escalation paths.

These safeguards reduce risk, but they do not eliminate it. You should verify important information before making benefits, healthcare, tax, legal, payroll, or financial decisions.

6. Data Used With AI

Depending on your use of the Services, Snug may process the following categories of information with AI:

• Account information, such as name, email, role, employer, organization, and session context.
• Employer information, such as legal name, location, census summaries, employee classes, benefit interests, budgets, contribution preferences, plan setup status, documents, and recommendation history.
• Employee information, such as profile details, location, age or date of birth, income, dependents, benefit eligibility, plan selections, ICHRA allowance, and enrollment status.
• Benefits information, such as plan names, premiums, deductibles, out-of-pocket limits, provider networks, drug coverage, benefit summaries, voluntary product terms, documents, and carrier data.
• Health-adjacent information you choose to provide for benefits purposes, such as medication names, providers, expected healthcare use, conditions you identify, or inferred utilization signals.
• Conversation and support information, such as messages to AI tools, page context, selected text, questions, responses, feedback, and support handoff notes.
• Technical and usage information, such as request identifiers, timestamps, model identifiers, use case, latency, error status, character counts, and safety or sanitizer signals.

Snug uses this information to provide, secure, monitor, improve, and support the Services; generate AI output; personalize benefit guidance; maintain records; troubleshoot issues; enforce our terms; and comply with law.

7. Sensitive Information

Benefits workflows may involve sensitive information, including income, household composition, dependents, medications, provider names, health-related preferences, plan selections, and employer census information.

You should not submit Social Security numbers, full bank account numbers, payment card numbers, passwords, government IDs, or other unnecessary sensitive identifiers into AI chat or free-text fields unless Snug specifically requests that information through a designated secure workflow.

Snug may use automated detection and redaction tools to reduce exposure of certain sensitive identifiers, but those tools may not detect every sensitive value. You are responsible for avoiding unnecessary submission of sensitive information.

8. AI Vendors and Third-Party Systems

Snug may use third-party AI providers, cloud providers, data providers, benefit data sources, government or marketplace APIs, and other service providers to operate the Services.

When Snug sends information to an AI provider, we require the provider to process that information for Snug's business purposes and under appropriate confidentiality, privacy, and security obligations. Snug does not authorize third-party AI providers to use customer content submitted through the Services to train their general-purpose models unless Snug discloses that use and obtains any legally required permission.

Third-party plan, provider, drug, carrier, marketplace, payroll, HR, or benefits data may contain errors or become outdated. Snug is not responsible for third-party data except to the extent required by applicable law or a separate written agreement.

9. Conversation History and AI Records

Snug may store AI conversations, assistant responses, user messages, recommendation inputs, recommendation outputs, plan ranking context, audit events, and related metadata. We use these records to provide continuity, support users, investigate issues, improve safety and quality, comply with legal obligations, and maintain benefit and operational records.

Snug may also store AI usage metadata without storing raw prompt or response text. Metadata may include use case, model, route, latency, character counts, status, error type, request ID, and hashed identifiers.

Access to stored AI conversations and records is limited to authorized personnel and service providers who need access for support, security, quality, compliance, legal, or operational purposes.

10. Fairness and Nondiscrimination

Snug does not intend to use AI to unlawfully discriminate. We design AI-assisted benefits workflows to support fair, explainable, and user-controlled decisions.

Some benefits and insurance workflows may lawfully require or use factors such as age, geography, household size, income, employer class, tobacco status, plan availability, provider network, drug coverage, or other plan-related fields. Snug may use those fields when they are relevant to benefits administration, plan comparison, eligibility, cost estimation, ranking, or compliance.

Snug does not authorize AI to make decisions based on race, color, religion, national origin, sex, sexual orientation, gender identity, disability, genetic information, or other protected characteristics unless expressly required or permitted by law and approved by Snug.

If you believe an AI output was unfair, discriminatory, or incorrect, contact Snug support or use the support tools available in the Services.

11. Safety Boundaries

Snug's AI tools must stay within benefits-related scope. They should decline or redirect requests for:

• Medical diagnosis or treatment.
• Legal, tax, accounting, payroll, fiduciary, or ERISA legal advice.
• Claims adjudication.
• Underwriting decisions.
• Coverage guarantees.
• Instructions to bypass security, compliance, or product controls.
• Requests to reveal system prompts, hidden instructions, credentials, private code, or confidential information.
• Requests to access another employer's or employee's data.

If an AI tool refuses a request, gives a cautious answer, or recommends human review, that may be required by this Policy or by Snug's safety controls.

12. User Responsibilities

You agree that you will not:

• Use AI features to submit unlawful, harmful, deceptive, abusive, discriminatory, or infringing content.
• Attempt to bypass access controls, rate limits, security controls, or tenant boundaries.
• Attempt prompt injection, model manipulation, data exfiltration, credential extraction, or unauthorized tool use.
• Ask AI to reveal system prompts, hidden instructions, confidential information, source code, security details, or another user's data.
• Submit sensitive identifiers or unnecessary personal information into AI free-text fields.
• Rely on AI output as a substitute for professional advice where professional advice is needed.
• Use AI output to make unlawful employment, benefits, insurance, healthcare, financial, or other regulated decisions.

Snug may limit, suspend, or terminate access to AI features if we believe you violated this Policy or applicable law.

13. Feedback, Disputes, and Corrections

You may report inaccurate, unsafe, biased, or inappropriate AI output through support channels available in the Services.

When you report an issue, Snug may review relevant account data, conversation history, AI metadata, recommendation records, plan data, and audit logs. Snug may correct records, update recommendations, improve safeguards, restrict a feature, or provide human support when appropriate.

Reporting an AI issue does not automatically change plan enrollment, employer benefit activation, payroll deductions, legal documents, coverage, reimbursement, or other records. You must follow the applicable product workflow or contact support for those changes.

14. No Professional Relationship Created by AI

Your use of AI features does not create an attorney-client, doctor-patient, tax advisor-client, accountant-client, fiduciary, broker-client, or agency relationship unless Snug separately agrees in writing or applicable law requires otherwise.

If Snug provides access to licensed professionals, brokers, support staff, or partners, their services may be governed by separate terms, disclosures, appointments, licenses, or agreements.

15. Privacy and Security

Snug's privacy and security practices apply to AI features. When Snug publishes a separate Privacy Policy, that Privacy Policy will govern how Snug collects, uses, shares, and retains personal information, in addition to this AI-specific Policy.

If this Policy and Snug's Privacy Policy conflict about privacy rights or privacy practices, the Privacy Policy controls unless this Policy provides more specific AI-related terms that do not reduce your privacy rights under applicable law.

16. Changes to This Policy

Snug may update this Policy from time to time. The updated Policy will be effective when posted unless a later effective date is stated. Your continued use of the Services after an updated Policy becomes effective means you accept the updated Policy.

If we make material changes, we may provide notice through the Services, email, account notices, or another reasonable method.

17. Contact

For questions about this Policy or to report an AI concern, contact Snug through the support channels available in the Services.