Automation Is No Defense
The RentGrow case answers the title’s question through a chain of ordinary decisions. A tenant-screening report begins with sources and identifiers, but it does not end there. Someone must decide which records belong to the applicant, whether several entries describe one proceeding, whether a disposition or legal restriction changes the meaning of an event, how a dispute is routed, and what the landlord or property manager is told after the report changes. This final synthesis connects those decisions and gives readers a practical accountability test: identify the handoff, inspect the evidence and status it preserves, follow the correction path, and keep the documents’ limits in view.
“Automation is no defense” does not mean that a computer must never assist with tenant screening. It means that automation does not answer the questions that make a housing report accurate. A system can combine information from multiple sources and apply matching or screening rules repeatedly, yet still count one case twice, use a weak identifier, omit a disposition, retain a legally restricted record, or close a dispute before the relevant evidence reaches the right agency. The result may be efficient and still be misleading.
The report is a chain of accountable choices
RentGrow occupies the center of the relationship described by the complaint. Upstream agencies, public-record vendors, and jurisdictions supply information. RentGrow allegedly assembles and evaluates that information in Tenant Screening Reports, including reports produced through ScreeningWorks Pro. A landlord or property manager uses the report to help determine whether an applicant meets rental criteria. The applicant is the person whose identity and history are being evaluated and who may need to inspect or challenge the report.
Each party sees a different part of the chain. A government recordkeeper may maintain a filing or judgment. A vendor may package information from one or more sources. RentGrow may use historical addresses, names, middle names, or initials to help vendors search and match records. A landlord may receive a report and an eligibility signal rather than the source history behind every line. The applicant may receive a file disclosure or dispute result that does not show the same information the landlord received. Accountability becomes difficult when those views are treated as interchangeable.
The first choice is identity. The complaint alleges that RentGrow used LexisNexis Accurint address-history and name information to locate additional addresses and supplement missing middle names or initials, then provided that information to vendors that searched for criminal and eviction records. This kind of enrichment can be useful: a limited public record may be difficult to find or connect to an applicant without an older address or fuller name. But an identifier that helps locate a possible record is not proof that the record belongs to the applicant. The more hidden the identifier, the harder it is for the person affected to test that connection.
The second choice is representation. The complaint alleges that RentGrow reported duplicate criminal and eviction information and presented stages of proceedings as separate records. The CFPB advisory opinion explains why that matters: a report should not make one conviction or one eviction appear to have occurred more than once, and it should keep the stages of one proceeding visibly connected. A report can therefore be inaccurate even when each individual source fragment is real. The error lies in the relationship the report communicates among the fragments.
The third choice is status. The CFPB says a report should include available disposition information when it reports arrests, criminal charges, eviction proceedings, or other court filings. It also says that records expunged, sealed, or otherwise legally restricted from public access should not be included when the report user could not obtain them directly from the government recordkeeper. Reporting periods add another boundary: for certain adverse information, the period generally runs from the adverse event and is not restarted by a later event.