Generative AI Prompt Use Policy Template for HIPAA-Compliant Medical Billing Companies

Kevin Henry

HIPAA

Published Updated

9 minutes read
Share this article
Generative AI Prompt Use Policy Template for HIPAA-Compliant Medical Billing Companies

Template notice: Adapt this policy to your organization’s HIPAA role, client agreements, applicable payer rules, and security program. Approval of an AI tool does not, by itself, establish HIPAA compliance; your organization remains responsible for its own obligations and contractual commitments.

What uses of generative AI are permitted or prohibited in medical billing?

Policy: Staff may use generative AI for approved revenue-cycle management tasks only when the tool and use case appear on the company’s approved-tool list and comply with this policy. Permitted examples may include drafting generic communications, summarizing non-sensitive public guidance, creating training materials from synthetic examples, and organizing de-identified workflow information.

Staff must not enter PHI or confidential client information into a tool or workflow that has not been approved for that data and purpose. Do not use AI to invent, alter, or embellish medical-record facts; create unsupported diagnoses or services; select codes solely to increase reimbursement; or submit, appeal, or communicate a claim without the required human review. AI output is a work aid, not a source of clinical documentation or payer authority.

Under this template, any use involving PHI requires explicit approval under the controls below. That is an internal safeguard—not a claim that HIPAA categorically prohibits every use of cloud-based AI. HHS explains that a cloud service creating, receiving, maintaining, or transmitting ePHI on behalf of a covered entity or business associate may be a business associate and that a BAA and other HIPAA safeguards are required in that situation. See HHS guidance on HIPAA and cloud computing.

Who approves AI tools and maintains the approved-tool list?

Policy: The Privacy Officer and Security Officer must jointly approve each AI tool and each proposed use case before staff use it for company work. They must obtain input from legal counsel, compliance, IT/security, and the billing or coding lead when the proposed use affects their responsibilities. A manager, employee, or vendor may not approve their own exception.

The request must identify the business purpose, users, data types, workflow, expected outputs, integrations, and whether PHI or other confidential information will be processed. The Security Officer documents the HIPAA Security Rule risk analysis and risk-management decision for relevant ePHI, including identified threats, safeguards, residual risks, and any required remediation. HHS describes risk analysis as an assessment of risks and vulnerabilities to ePHI and a foundational part of Security Rule compliance; see HHS guidance on risk analysis.

The Security Officer maintains the approved-tool list and records each tool’s approved purpose, permitted data, configuration, accountable owner, review date, and any restrictions. Reassess approval when the vendor changes its data practices, model, hosting, subcontractors, or security controls, or when the company changes the workflow. Suspend use while a material risk or suspected incident is investigated.

How should staff handle PHI in prompts?

Policy: Before entering information into an approved tool, staff must confirm that the use is authorized, the tool is approved for that information, and the prompt contains only the minimum information needed for the task. Where practical, use synthetic or de-identified examples instead of patient-specific information. Removing a name alone may not de-identify information if other details could identify the person.

Never include PHI in public or personal AI accounts, browser extensions, unapproved meeting assistants, or tools accessed with personal credentials. Do not paste full medical records, claim files, remittance advice, or client data into a prompt when a limited excerpt or non-identifying description will suffice. Avoid including direct identifiers and account credentials unless the approved workflow specifically requires them.

Use company-managed accounts and approved access controls. Do not share logins, move prompts or outputs to personal storage, or use AI tools to bypass client access restrictions. Report a suspected disclosure, accidental prompt submission, or unusual AI output through the company’s incident-reporting process promptly. The minimum necessary standard calls for reasonable efforts to limit PHI use, disclosure, and requests to what is needed for the purpose; see HHS guidance on the minimum necessary requirement.

What vendor agreements and security reviews are required?

Policy: Before an AI vendor receives or processes PHI, Privacy, Security, and Legal must determine whether the vendor is acting as a business associate or business associate subcontractor for the proposed workflow. If the vendor creates, receives, maintains, or transmits PHI on behalf of the company or its client in a role that requires a BAA, execute the required agreement before use. A vendor’s marketing statement or general privacy policy is not a substitute for the required agreement.

Review the BAA and related terms for permitted uses and disclosures, safeguards, incident and breach reporting, subcontractors, retention and deletion, return of data, and restrictions on using prompts or outputs to train or improve models. Confirm that the vendor’s commitments match the actual product configuration and your client contracts. HHS describes required BAA provisions and downstream subcontractor agreements in its Business Associates FAQs and sample business associate agreement provisions.

Before approval, assess the vendor’s security controls and the end-to-end workflow, including data storage and transmission, access controls, authentication, audit logging, retention, deletion, incident response, integrations, and administrative access. Document the review and any mitigations in the company’s risk-management records. HHS identifies access controls and audit controls among the Security Rule’s technical safeguards; see the HHS summary of the HIPAA Security Rule. A vendor’s security report or claimed certification may inform review, but it does not establish that the company or the tool is “HIPAA certified.”

Ready to simplify HIPAA compliance?

Join thousands of organizations that trust Accountable to manage their compliance needs.

Get Started Free Today

How should AI-generated billing and coding work be reviewed?

Policy: A designated, qualified employee must independently check AI-generated billing, coding, coverage, denial, appeal, and documentation-related content before it is relied on, communicated, or used in a claim. The reviewer must compare the output with the source record, current code sets, applicable payer rules, client instructions, and relevant documentation. AI output must not add facts that are absent from the record or resolve an ambiguity by guesswork.

For coding or claim recommendations, the reviewer must verify code validity, specificity, modifiers, units, dates, eligibility, and supporting documentation as applicable to the task. Escalate uncertain or high-impact matters to the coding or compliance lead, and obtain provider clarification through approved channels when the record does not support a decision. Do not let AI independently finalize or transmit a claim unless a separately approved workflow expressly permits the specific action and includes appropriate controls.

Record the reviewer, review date, material corrections, and the source information used when required by the workflow or client contract. CMS states that correct coding is key to valid claims and directs users to use current valid codes at the highest available specificity; its documentation guidance also explains that claims must be supported by medical-record documentation. See CMS guidance on correct coding and CMS guidance on medical-record documentation.

What training and oversight should the policy require?

Policy: Before using an approved AI tool, staff must complete role-appropriate instruction on this policy, PHI handling, approved tools and use cases, prompt safety, output verification, incident reporting, and relevant client-specific requirements. Managers must ensure that training reflects each employee’s access and responsibilities, including coding, billing, appeals, and quality review.

Provide additional instruction when tools, workflows, policies, or security risks change, and set a refresher schedule appropriate to the organization’s risk and employer-specific requirements. HIPAA requires a security awareness and training program for workforce members, but does not set a universal annual certificate-renewal schedule. A training completion certificate documents completion of a course; it does not certify that the organization or its AI workflow is HIPAA-compliant. See HHS’s HIPAA audit protocol on security awareness and training.

The Privacy Officer, Security Officer, and process owners must periodically review the approved-tool list, usage records, access permissions, audit logs, incidents, and quality findings. Define in company procedures how often reviews occur, how exceptions are approved, and how concerns are escalated. Apply corrective action when staff bypass approval, disclose information improperly, or rely on unverified output.

Conclusion

A practical generative AI policy for medical billing sets clear permitted uses, requires advance tool and workflow approval, protects PHI in prompts, reviews vendor agreements and security risks, and keeps qualified people accountable for billing and coding decisions. Treat this template as a starting point and align it with your organization’s HIPAA obligations, client agreements, and documented risk analysis.

FAQs.

Can medical billing staff enter PHI into a public generative AI tool?

Not under this template unless the specific tool and use case have been formally approved for PHI and the required safeguards and agreements are in place. A public or consumer-facing tool should be treated as unapproved by default. If a vendor processes ePHI on behalf of a covered entity or business associate, HHS guidance indicates that a BAA and compliance with applicable HIPAA requirements are needed. See HHS guidance on HIPAA and cloud computing.

When is a business associate agreement required for an AI vendor?

A BAA is generally required when the vendor acts as a business associate by creating, receiving, maintaining, or transmitting PHI on behalf of a covered entity or business associate, subject to HIPAA’s definitions and exceptions. If a vendor’s subcontractor performs such work, the business associate must obtain the required downstream agreement. Determine the vendor’s role for the actual service and workflow before sending PHI; see HHS Business Associates FAQs.

Who should approve generative AI tools for billing workflows?

Under this template, the Privacy Officer and Security Officer jointly approve the tool and use case, with input from Legal, Compliance, IT/security, and the billing or coding lead as needed. Approval should follow documented review of the intended purpose, data, vendor terms, security controls, and relevant risks—not merely a department manager’s informal permission.

Must staff verify AI-generated billing or coding content before use?

Yes, under this policy. A designated employee must compare the output against the source record, current code sets, payer rules, and applicable documentation before relying on it or using it in a claim. CMS guidance emphasizes correct, current coding and documentation that supports billed services; AI-generated content should not replace that verification.

Share this article

Ready to simplify HIPAA compliance?

Join thousands of organizations that trust Accountable to manage their compliance needs.

Get Started Free Today

Related Articles