Fairmark

Housing providers and marketplaces increasingly depend on automated assistants for lead handling, recommendations, screening-adjacent workflows and resident communication. Vendors need a credible way to show how those systems behave under Fair Housing scenarios, while buyers and insurers need evidence they can inspect. The supplied research found individual training credentials and a responsible-AI credential, but not the exact marketplace-software assessment proposed here.

Fairmark defines a controlled assessment program rather than selling an instant compliance badge. An applicant declares product version, intended uses, prohibited uses, data inputs, markets and buyer controls. A versioned test pack exercises protected-class proxies, steering, differential treatment, accessibility, escalation and monitoring cases. Results become evidence candidates for an independent reviewer; they do not prove lawful operation in every deployment.

Assessment completion, reviewer decision, mark license, marketplace acceptance, insurance underwriting and customer deployment remain separate. The mark has a published scope, expiry, surveillance requirements and a correction or withdrawal process. An insurer may consume evidence only under a separate relationship and makes its own underwriting decision.

The first release should assess one narrow agent category in one market and avoid tenant selection or other high-consequence decisions. Its defensibility depends on transparent governance, repeatable evidence and trusted review, not on claiming regulatory authority.

Risk, compliance or product leader at a housing-software vendor or real-estate marketplace seeking reviewable evidence for an AI agent