Lesson objective
Use the NIST AI RMF functions, the Generative AI Profile's risks, an AI management system and the NIST CSF as voluntary structure for a brokerage programme, and set a review cycle that keeps it current.
Plain-language explanation
Why this matters now
A brokerage programme written only from state rules and regulator guidance can end up as a list of controls with no shape. Voluntary frameworks supply the shape: a way to check that nothing important is missing and a vocabulary that insurers, larger clients and technology partners recognise. They are also easy to misuse, by presenting them as law or claiming a status the brokerage does not have.
Four frameworks are useful here. Each is voluntary for a brokerage.
- Best practice (NIST AI Risk Management Framework, AI RMF 1.0, NIST AI 100-1, released January 26, 2023; voluntary): NIST describes it as "voluntary, rights-preserving, non-sector-specific, and use-case agnostic." Its Core has four functions: GOVERN, MAP, MEASURE and MANAGE. NIST's web page states that "The AI RMF 1.0 is being revised as part of the White House AI Action Plan"; as of September 15, 2026 no revised version had been published. Date every reference to 1.0.
- Best practice (NIST AI 600-1, Generative AI Profile, released July 26, 2024; voluntary): a cross-sectoral companion to the AI RMF that defines 12 risks unique to or made worse by generative AI.
- Best practice (ISO/IEC 42001:2023, published December 2023, Edition 1; voluntary unless adopted by contract or by an organisation seeking conformity): ISO describes it as specifying "requirements for establishing, implementing, maintaining, and continually improving an Artificial Intelligence Management System (AIMS)," using the Plan-Do-Check-Act method, for "organizations of any size involved in developing, providing, or using AI-based products or services." The standard's text is paid; this lesson relies only on ISO's public description and quotes no clauses.
- Best practice (NIST Cybersecurity Framework 2.0, February 26, 2024; voluntary): six functions, GOVERN, IDENTIFY, PROTECT, DETECT, RESPOND and RECOVER. It "does not prescribe how outcomes should be achieved." Its GOVERN function includes cybersecurity supply chain risk management (category GV.SC).
Professional guidance (NAR overview page; voluntary): NAR offers an AI Policy Template for Brokers to members. It was not reviewed for this programme; treat it as a member resource.
A question to sit with
An insurer's questionnaire asks whether your brokerage "follows a recognised AI risk framework." What could you truthfully say, and what would you have to be careful not to claim?
What goes wrong
Framework as law. A policy says "NIST requires brokerages to..." NIST requires nothing of anyone; the framework says so itself.
The certificate that does not exist. A brokerage's website says it is "NIST AI RMF certified." There is no NIST certification for the AI RMF.
Borrowing a revised version that does not exist. Someone repeats what the "new" AI RMF is expected to say. No revised version has been published; describe only 1.0.
Quoting what you have not read. A policy cites "ISO 42001 clause 6.1" copied from a vendor's brochure. If the firm has not bought and read the standard, it should not quote its clauses.
Framework theatre. The programme is reorganised under GOVERN, MAP, MEASURE and MANAGE headings, and nothing about the actual controls changes.
The practical answer
Map the programme to the four AI RMF functions. NIST says GOVERN "applies to all stages," while MAP, MEASURE and MANAGE can be applied to specific systems and lifecycle stages, and that after GOVERN outcomes are in place "most users of the AI RMF would start with the MAP function." A brokerage mapping:
| AI RMF 1.0 function | Where it lives in the brokerage programme |
|---|---|
| GOVERN | Owner, named-owner register, decision rights, stop authority, policy, training |
| MAP | Permitted-use inventory: each use's task, data, tier and who is affected |
| MEASURE | Review records, outcome audits, fair housing sampling of reach and treatment |
| MANAGE | Decisions on audit findings: change, restrict, suspend or withdraw a use |
If a function has nothing under it, the programme has a gap. That is the test worth running; the headings themselves add nothing.
Check the trustworthiness list against your inventory. AI RMF 1.0 lists the characteristics of trustworthy AI: valid and reliable; safe; secure and resilient; accountable and transparent; explainable and interpretable; privacy-enhanced; and fair with harmful bias managed. For each public or client-facing use, ask which characteristic is most at risk and which control addresses it.
Choose the GenAI Profile risks that fit a brokerage. The profile lists 12 risks. Several map directly to brokerage work; others rarely arise.
| GenAI Profile risk | Brokerage example | Programme component |
|---|---|---|
| Confabulation | Invented features, wrong square footage, a rule that does not exist | Review against source; audit |
| Data Privacy | Client identifiers pasted into a public tool | Inventory data rules |
| Harmful Bias or Homogenization | Chatbot answers that differ by neighbourhood or by who is asking | Fair housing review |
| Human-AI Configuration | Automation bias in reviewers | Review evidence; audit; training |
| Information Integrity | Altered listing images; synthetic media | Advertising review; Module 09 |
| Information Security | Tool compromise; exposed accounts | Security programme; Module 08 |
| Intellectual Property | Photos or descriptions used without permission | Advertising review; contracts (Module 07) |
| Value Chain and Component Integration | A vendor's tool built on another provider's model | Supplier assessment; Module 05 |
Eight of the twelve, and every row points to a component that already exists or a module that builds one.
Consider a management-system approach as the firm grows. ISO describes ISO/IEC 42001 as a management system standard that, "rather than looking at the details of specific AI applications," manages AI risks and opportunities across an organisation using Plan-Do-Check-Act. For a multi-office brokerage, the idea is useful even without pursuing conformity: plan (the programme), do (the inventory and review), check (the audit) and act (the changes). Whether to buy the standard, seek conformity or do neither is a business decision; nothing in this module's sources requires it.
Borrow the CSF for the supplier and security questions. CSF 2.0's GOVERN function includes supply chain risk management (GV.SC). A brokerage's AI tools are almost all bought, so the programme's supplier assessment (Module 05) and security controls (Module 08) can use CSF 2.0 vocabulary. Again, the CSF is voluntary.
Set the review cycle. AI Coalition Network recommendation: review the whole programme at least once a year, and review the affected part within a set period when any of these happen:
- a state rule or regulator guidance that the programme cites changes, is withdrawn or is replaced;
- a tool on the inventory changes materially, including a model change announced by the vendor;
- an audit finds a pattern of errors, or a complaint traces back to an AI use;
- the owner, a managing broker or the compliance coordinator changes;
- NIST publishes a revised AI RMF.
In Texas, the duty to maintain current written policies means the cycle is also how the broker shows the policies are current.
Remember this
- NIST AI RMF 1.0, the GenAI Profile, ISO/IEC 42001 and CSF 2.0 are voluntary for a brokerage; none is a legal requirement.
- There is no NIST certification; do not claim one, and do not describe a revised AI RMF that has not been published.
- Map the programme to GOVERN, MAP, MEASURE and MANAGE to find gaps, not to rename headings.
- Pick the GenAI Profile risks that fit brokerage work and tie each to a component.
- Review the programme yearly and on five triggers, including a rule change, a tool change and a change of owner.
Real estate example
Silverpine Properties is a fictional 110-agent brokerage with offices in two states. Its errors-and-omissions insurer sent a new questionnaire asking whether the firm "follows a recognised AI risk framework (e.g., NIST, ISO)." The marketing director drafted an answer: "Silverpine is fully NIST and ISO compliant." Managing broker Eleanor Voss stopped it. There is no NIST certification for the AI RMF, the firm had never bought or read ISO/IEC 42001, and "compliant" implied a status nobody had assessed. Eleanor asked the compliance lead, Hugo Martel, to map the existing programme against AI RMF 1.0. Under GOVERN he found the owner, the named-owner register and the decision-rights table. Under MAP, the permitted-use inventory. Under MEASURE, review records and the quarterly audit. Under MANAGE he found almost nothing: audit findings were reported but no record showed what the firm had decided to do about them. That gap was real, and the framework had found it. Hugo added a findings log with a decision, an owner and a date for each finding. He then took the GenAI Profile's 12 risks and marked which applied to Silverpine's uses; eight did, and each was tied to a component. The insurer's answer read: "We organise our AI programme using NIST AI RMF 1.0 (January 2023) and NIST AI 600-1 (July 2024), voluntary frameworks. We hold no certification to them, as none exists. Our programme is owned by our managing broker and was last audited on 30 June." Eleanor added a review trigger for any revised AI RMF NIST publishes.
Responsible-use warning
Do not describe any of these frameworks as a legal requirement, claim a NIST certification, quote ISO clauses the firm has not read, or describe the contents of a revised AI RMF that has not been published. Insurers, clients and investigators can check each of those claims, and an overstatement in one sentence undermines the credibility of the rest. Use the frameworks to find gaps, and describe them exactly, with their dates and their voluntary status.
Human-review requirement
The designated, responsible or supervising broker owns the programme's structure and any public or insurer statement about frameworks, and approves the wording before it is sent. The compliance lead maps the programme to the frameworks and records the gaps found. The broker checks that every framework is described as voluntary and dated, that no certification is claimed, that no unread clause is quoted, and that the review cycle runs yearly and on each trigger.
Practical exercise
Use a fictional brokerage that already has an inventory, a decision-rights table and a quarterly audit.
- Map each existing programme component to GOVERN, MAP, MEASURE or MANAGE, as NIST AI RMF 1.0 describes them.
- Mark any function with nothing under it, and write the control that would fill the gap.
- For three public or client-facing uses, name the trustworthiness characteristic most at risk and the control that addresses it.
- Go through the GenAI Profile's 12 risks and mark which apply to your firm's uses. For each that applies, name the component or module that addresses it.
- Write two sentences on whether the firm will use a management-system approach such as ISO/IEC 42001, labelled as a business decision.
- Draft the firm's answer to an insurer asking whether it follows a recognised AI risk framework, avoiding any claim of certification or compliance.
- Write the review cycle: the annual date and the five triggers, with who acts on each.
- List every sentence in your current policy that mentions NIST, ISO or NAR, and correct any that overstates the source's status or date.
- Note the date you will next check NIST's AI RMF page for a revised version.
- Name the person who approves any public, client or insurer statement about frameworks, and write the one rule they apply before approving it.
Knowledge check
-
What are the four functions of the NIST AI RMF 1.0 Core, and which applies to all stages?
Answer guidance: GOVERN, MAP, MEASURE and MANAGE. GOVERN applies to all stages; MAP, MEASURE and MANAGE can be applied to specific systems and lifecycle stages. The framework is voluntary and dated January 2023.
-
Can a brokerage be certified to the NIST AI RMF?
Answer guidance: No. There is no NIST certification for the AI RMF, and the framework is voluntary. A brokerage may say it uses AI RMF 1.0 to organise its programme, but not that it is certified or compliant.
-
What does ISO say ISO/IEC 42001:2023 specifies, and what may this lesson not do with it?
Answer guidance: Requirements for establishing, implementing, maintaining and continually improving an AI management system, using Plan-Do-Check-Act. The lesson relies only on ISO's public page and quotes no clauses, because the text is paid and was not reviewed.
-
Name three triggers for reviewing the programme outside the annual cycle.
Answer guidance: A change or withdrawal of a cited rule or guidance, a material change to a tool on the inventory, an audit pattern or complaint traced to AI, a change of owner or key role, or a revised AI RMF being published.
Sources and further reading
- National Institute of Standards and Technology — Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (2023-01-26). Accessed 2026-09-15.
- National Institute of Standards and Technology — Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1 (2024-07-26). Accessed 2026-09-15.
- International Organization for Standardization (ISO) and International Electrotechnical Commission (IEC), ISO/IEC JTC 1/SC 42 — ISO/IEC 42001:2023 Information technology — Artificial intelligence — Management system (2023-12-18). Accessed 2026-09-15.
- National Institute of Standards and Technology — The NIST Cybersecurity Framework (CSF) 2.0 (2024-02-26). Accessed 2026-09-15.
- National Association of REALTORS® — Artificial Intelligence (AI) in Real Estate. Accessed 2026-09-15.
- Texas Real Estate Commission — 22 Texas Administrative Code §535.2, Broker Responsibility. Accessed 2026-09-15.