Compliance Guide | Healthcare

    HTI-5 and Health Plans Building Prior Authorization Agents: Applicability and Runtime Evidence

    As of November 2026, HTI-5 is a final rule that streamlines the ONC Health IT Certification Program, and its requirements are written for health IT developers. A health plan that builds its own LangGraph prior authorization agent and does not develop certified health IT is not the rule's primary target. Exposure may arise indirectly through certified vendor components. HIPAA Security Rule safeguards apply to the agent whatever the HTI-5 answer is, so runtime policy enforcement and durable evidence are worth building either way. This guide is not legal advice; counsel should sign off on the determination.

    Where policy can be enforced in a LangGraph agent

    1. 1

      Tool calls

      Decide whether the agent may invoke the tool with these arguments. Evidence: tool, arguments scope, policy version, outcome.

    2. 2

      PHI retrieval

      Limit data scope and redact fields not needed for the task. Evidence: data scope accessed, fields redacted, outcome.

    3. 3

      Node transitions

      Identify transitions that can influence a coverage determination and route them to review. Evidence: transition, trigger, route taken.

    4. 4

      Human interrupts

      Pause for approval before consequential steps. Evidence: approver identity, timing, decision, any override.

    5. 5

      Graph state

      Keep PHI out of state that is passed onward or retained without need. Evidence: what state held at decision points.

    1. Complete and record the applicability determination
    2. Inventory tools and ePHI-returning data sources
    3. Confirm certification status of vendor components
    4. Route coverage-influencing transitions to human review
    5. Version policy and retain evidence durably

    Criteria for evaluating runtime controls and evidence

    • Can you show, for any run, which policy was in force and whether each tool call was allowed, redacted, blocked or escalated?
    • Is enforcement in the execution path, or does the system only observe afterward?
    • How long is evidence retained for PHI retrievals, and how is it protected from alteration?
    • Are human approvals captured with approver identity, timing and outcome?
    • Can evidence be exported and mapped to HIPAA Security Rule safeguards and reviewed by counsel against CMS requirements?
    • Are policy changes approved and versioned, so past behavior ties to the policy in force at the time?

    Who HTI-5 regulates

    HTI-5 is ONC's deregulatory final rule. ONC states that it reduces burden on health IT developers by streamlining the Certification Program, removing and revising requirements, and resetting the program's scope as a foundation for FHIR-based API requirements. Its Real World Testing and Standards Version Advancement Process changes sit under the heading for health IT developers. The rule is about certification of health IT modules, not about how a payer runs its internal decision systems.

    The rule's AI content is framed around AI-enabled interoperability through modernized standards and certification. It does not describe agent governance. Nothing in the rule's stated purpose creates a requirement that a plan govern an in-house agent in a particular way.

    Timing varies by criterion. Some changes take effect on the final rule's effective date, family health history is removed effective January 1, 2027, and the transitions of care revision takes effect January 1, 2030. HTI-5 also retained, unchanged, 39 certification criteria, including those adopted in the HTI-4 electronic prior authorization and real-time prescription benefit rule. That retention matters if your agent connects to certified electronic prior authorization components.

    Three applicability outcomes for a LangGraph agent

    Direct applicability is plausible only if the plan, or an affiliate, develops health IT that is certified under the ONC program. In that case the certified module, not the internal agent as such, is what the certification requirements address. The plan's own certification and Conditions and Maintenance of Certification obligations would be the starting point.

    Indirect applicability is the more common case. The agent calls certified EHR, API or electronic prior authorization components from vendors. The plan's exposure is then contractual and operational: which certified capabilities the agent depends on, what the vendor commits to, and how criteria changes affect those components over time.

    No applicability is a legitimate outcome when the plan neither develops nor depends on certified modules in the agent path. Record that determination, the facts behind it and the counsel sign-off, and revisit it when architecture or vendors change. The applicability analysis should not be confused with the controls question. Certification criteria such as auditable events describe certified modules and are a different construct from controls on an in-house agent, so they are a poor basis for designing agent audit controls.

    HTI-5 compared with adjacent obligations

    These regimes answer different questions. Treating them as one checklist tends to hide gaps, so assess each on its own terms.

    • HTI-5: ONC Certification Program rule aimed at health IT developers. Concerns certification criteria, real world testing and related program conditions.
    • HTI-4 criteria: Electronic prior authorization and real-time prescription benefit certification criteria, retained unchanged by HTI-5. Relevant when the agent depends on certified components.
    • HIPAA Security Rule: Applies to ePHI created, received, used or maintained by a covered entity or business associate. Requires administrative, physical and technical safeguards, including audit controls under 164.312(b).
    • CMS interoperability and prior authorization requirements: A separate regime with its own obligations and dates, assessed independently of HTI-5 and with counsel.
    • HIPAA Security Rule NPRM (January 2025): A proposal to strengthen the Security Rule. As of November 2026 it should be tracked, not treated as binding.

    What audit-ready evidence looks like

    HIPAA 164.312(b) calls for mechanisms that record and examine activity in systems that contain or use ePHI. For an agent, the practical question an auditor or regulator will ask is what the agent was permitted to do and what it actually did. Answering both requires two linked records: the policy in force at the time, and the outcome of each enforced action.

    Per run, capture the agent identity, the policy version, each tool invoked, the data scope accessed, the enforcement outcome and any human approver. Keep this evidence durable and separate from the agent's own state so the agent runtime cannot alter it. These are design recommendations rather than regulatory text, and the mapping of each field to a specific safeguard should be confirmed with counsel against the primary regulation.

    There are tradeoffs. Blocking and escalation add latency and reviewer workload, and tightly scoped retrieval can reduce the context the agent has. Starting with the transitions that can influence a coverage determination keeps the review burden proportionate.

    How Trussed fits

    Trussed enforces policy in the execution path across models, agents, tools and MCP servers, at the level of individual interactions and tool calls. Actions are allowed, redacted, blocked, rate limited or escalated for human approval, enforced at runtime, not reviewed after the fact. That enforcement generates audit-ready evidence continuously, mapped to HIPAA among other frameworks.

    Trussed runs in production with healthcare organizations, and can be self-managed in the customer's cloud with configurable retention and data residency. It does not replace the applicability analysis, which remains a legal determination for the plan.

    Applicability and control path for a prior authorization agent

    1. Applicability

    Developer of certified health IT, user of certified modules, or neither.

    2. Adjacent obligations

    HIPAA Security Rule and CMS prior authorization requirements apply on their own terms.

    3. Runtime enforcement

    Allow, redact, block or escalate at each tool call and PHI retrieval.

    4. Evidence

    Per-run record of policy in force, action taken and human approver.

    Review your agent's enforcement points and evidence

    Walk through your prior authorization agent's tool calls and PHI retrievals and see how enforced policy produces audit-ready evidence.

    Explore Agent and Tool-Call Enforcement