# Colorado's AI Proposal Raises New Compliance Questions for Lenders
Colorado lawmakers are moving forward with AI regulation that could reshape how mortgage lenders operate, but the Mortgage Bankers Association warns the proposal contains vague language that creates compliance headaches for creditors across the state.
The bill focuses on automated decision-making technology (ADMT) and requires lenders to disclose when they use AI systems to evaluate loan applications, price mortgages, or make other credit decisions. Lenders must also explain the "consequential decisions" that result from these systems. The problem: nobody has defined what those terms actually mean in practice.
The MBA submitted formal comments pushing Colorado legislators to clarify the definitions before the bill becomes law. Without clear guidance, lenders face the risk of unintentional violations. A bank that believes it is compliant might discover regulators interpret the rules differently. That uncertainty drives up compliance costs and creates legal exposure.
Colorado joins a growing number of states tightening rules around AI in lending. New York adopted similar requirements last year. The Federal Reserve, Office of the Comptroller of the Currency, and Consumer Financial Protection Bureau have all issued guidance warning lenders about algorithmic bias and discrimination risks. But state-by-state rules create a patchwork that forces national lenders to build different compliance systems for different markets.
Here's what matters for market participants. Mortgage lenders operating in Colorado will need to review their underwriting systems and identify where AI plays a role. That includes obvious cases like algorithmic loan decisioning, but also less obvious applications: AI that predicts borrower default risk, systems that flag suspicious applications for fraud, or algorithms that price risk and determine interest rates.
For borrowers, the disclosure requirement means more paperwork in the loan process. Lenders will need to tell you if an algorithm affected your approval decision or interest rate. You gain the right to understand how machines evaluated your creditworthiness. Regulators expect this transparency to catch discriminatory outcomes before they cause harm.
For mortgage brokers and lenders, compliance teams will need to audit their technology stacks. Many lenders already use predictive analytics and machine learning without labeling it explicitly as ADMT. Those systems may fall under Colorado's definition once the legislature clarifies the language. The cost of compliance falls on lenders, not borrowers, but those costs eventually flow into loan pricing.
Sellers benefit indirectly from clearer disclosure rules. Homes sell faster when buyers obtain financing smoothly. Regulatory clarity accelerates the approval timeline instead of slowing it down. Delayed loan decisions hurt transactions.
The MBA's push for clearer definitions reflects a broader pattern. Regulators write rules; industry groups demand specificity; legislatures scramble to revise language. Colorado has an opportunity to avoid that cycle by getting the definitions right from the start.
The bill remains in committee. Lenders should monitor its progress closely. If Colorado passes vague language, expect regulators to interpret it broadly. That interpretation becomes the de facto standard for compliance. Better to influence the bill now than fight enforcement actions later.